WO2010080719A1 - Search engine for refining context-based queries based upon historical user feedback - Google Patents

Search engine for refining context-based queries based upon historical user feedback Download PDF

Info

Publication number
WO2010080719A1
WO2010080719A1 PCT/US2010/000075 US2010000075W WO2010080719A1 WO 2010080719 A1 WO2010080719 A1 WO 2010080719A1 US 2010000075 W US2010000075 W US 2010000075W WO 2010080719 A1 WO2010080719 A1 WO 2010080719A1
Authority
WO
WIPO (PCT)
Prior art keywords
query
user
word
words
keywords
Prior art date
Application number
PCT/US2010/000075
Other languages
French (fr)
Inventor
Yuheng Xie
Wenwu Ou
Original Assignee
Alibaba Group Holding Limited
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Alibaba Group Holding Limited filed Critical Alibaba Group Holding Limited
Priority to JP2011545381A priority Critical patent/JP5679993B2/en
Priority to EP10729396A priority patent/EP2382534A4/en
Publication of WO2010080719A1 publication Critical patent/WO2010080719A1/en

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90Details of database functions independent of the retrieved data types
    • G06F16/95Retrieval from the web
    • G06F16/951Indexing; Web crawling techniques
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33Querying
    • G06F16/3331Query processing
    • G06F16/3332Query translation
    • G06F16/3338Query expansion
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/23Updating
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/24Querying
    • G06F16/245Query processing
    • G06F16/2457Query processing with adaptation to user needs
    • G06F16/24573Query processing with adaptation to user needs using data annotations, e.g. user-defined metadata
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90Details of database functions independent of the retrieved data types
    • G06F16/95Retrieval from the web
    • G06F16/953Querying, e.g. by the use of web search engines
    • G06F16/9535Search customisation based on user profiles and personalisation
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90Details of database functions independent of the retrieved data types
    • G06F16/95Retrieval from the web
    • G06F16/953Querying, e.g. by the use of web search engines
    • G06F16/9538Presentation of query results

Definitions

  • the present application relates to data processing technology and, in particular, to a method and a device for word matching and information query.
  • Latent word meaning usually refers to the latent meaning of a word or phrase and it usually may be expressed in one or more words or phrases.
  • latent word meaning of "#KH” (fridge) generally refers to (refrigerator)
  • latent word meaning of "f&tf ⁇ " cotton slippers
  • ⁇ t ⁇ tfe ⁇ all-cotton slippers
  • the indexing mode of a search engine typically includes separate word search, word partitioning indexing, and hybrid indexing.
  • the distance between the separate words in a file typically needs to be calculated.
  • the efficiency is often low and accuracy often poor.
  • the problem is particularly pronounced in languages without natural word separators (e.g., spaces) between words, such as Chinese.
  • word separators e.g., spaces
  • the word partitioning search technique has higher accuracy and is fast, but the recall rate is often low.
  • FIG. IA is a block diagram illustrating an embodiment of a query processing system.
  • FIG. IB is a flowchart of an embodiment of a process for performing a query.
  • FIG. 2 is a block diagram illustrating an embodiment of a query processing system.
  • FIG. 3 is a block diagram illustrating an embodiment of the matching module.
  • FIG. 4 is a block diagram illustrating an embodiment of the first-query- keyword obtaining module.
  • FIG.5 is a flowchart illustrating an embodiment of a process for word matching.
  • FIG.6 is a block diagram illustrating an embodiment of a device for word matching.
  • the invention can be implemented in numerous ways, including as a process; an apparatus; a system; a composition of matter; a computer program product embodied on a computer readable storage medium; and/or a processor, such as a processor configured to execute instructions stored on and/or provided by a memory coupled to the processor.
  • these implementations, or any other form that the invention may take, may be referred to as techniques.
  • the order of the steps of disclosed processes may be altered within the scope of the invention.
  • a component such as a processor or a memory described as being configured to perform a task may be implemented as a general component that is temporarily configured to perform the task at a given time or a specific component that is manufactured to perform the task.
  • the term 'processor' refers to one or more devices, circuits, and/or processing cores configured to process data, such as computer program instructions.
  • FIG. IA is a block diagram illustrating an embodiment of a query processing system.
  • query processing system 150 includes a query processing server 152, which in some embodiments include one or more devices having one or more processors coupled to one or more memories.
  • Query processing server 152 includes one or more interfaces configured to receive user inputs, for example user inputs sent by client devices such as 154 over a network 155, which may be the Internet or other communications network. Examples of a communication interface include without limitation external connections, such as a port, cable, wired or wireless network interface card, etc., and internal connections, such as a communication bus.
  • the client device is a computing device such as a computer or mobile device with networking capabilities.
  • Query processing server 152 has access to a user feedback database 156 that stores user feedback logs.
  • the query processing server processes the user input to generate query keywords based on the user input and user feedback log.
  • the query keywords are sent to a search engine 158 which locates and returns search results based on the query keywords.
  • the search engine is configured to search for specific content stored in a database, such as product information on an electronic commerce website.
  • the search engine is configured as a general purpose search engine that searches for content on the Internet. The functions of the keyword search engine and the query processing server are combined in some embodiments; thus, the query processing server uses the keywords to perform search and returns the results directly.
  • the search results and user response to the search results (such as user selection/click-through) are updated in the user feedback database.
  • FIG. IB is a flowchart of an embodiment of a process for performing a query.
  • Process 100 may be performed on a system such as 150.
  • a user input that includes a plurality of symbols is received.
  • the user input is sent from a client device such as 154 via a web browser or other user interface tool to the query processing server.
  • a symbol refers to a character or a letter. A number of characters/letters form a word.
  • a first set of query keywords are determined based on the user input.
  • the first set of query keywords may include one or more words.
  • the user input is used directly as the first set of query keywords.
  • the user input is partitioned and the partitioned words are used as the first set of query keywords.
  • the user input is in a letter based language such as English, where words are naturally separated by spaces. The user input therefore does not require context based processing and the first set of query keywords include the words in the user input and/or stemmed input words.
  • character based languages such as Chinese is used as the user input language. Since there are no specific character dividers in such languages, the input symbols are partitioned into words based on context, and the first set of query keywords include the partitioned words.
  • the first set of query keywords also includes the characters in the user input, which are available without any partitioning.
  • a user feedback log is obtained based on the first set of query keywords.
  • the user feedback log which is obtained from a database such as 156 in some embodiments, includes historical query results using the first set of query keywords as the target (or object) of the query and the selection frequency (also referred to as the click- through rate) by the users.
  • the selection frequency may include selection frequency on the previous query results and/or selection frequency on the contents of previous results.
  • the user feedback log is gathered by the search engine.
  • keywords input by the users, past query results e.g., web page file identifiers/universal resource locators (URLs)
  • click frequencies on the past query results, display frequencies of the past query results, etc. are collected by the search engine and stored.
  • the first set of query keywords include the words "western medicine" or "MI5"
  • the user feedback log includes records of all or some of the results
  • a second set of query keywords that matches the first set of query keywords based on the user feedback log is determined.
  • the second set of query keywords includes latent meaning words derived based on the user feedback log. All or a portion (such as a randomly selected portion) of the previous records may be used to establish latent word meaning.
  • the automatic finding of latent words with latent meaning refers to the finding of one or more words (or phrases) with relevant or similar word meaning as the original word.
  • the latent word meaning relationship between the query word and the historical query results that embody the user's intention can be reliably and automatically found.
  • the latent word meaning relationship is in turn used to improve the accuracy rate and intelligence of the search engine. Therefore, the user feedback log may include all previous historical query results that use the first set of query keywords as the search target and all previous click selecting frequencies of users on the historical query results.
  • determining the second set of query keywords includes: obtaining the content of historical query results of the first set of keywords, partitioning the content of the historical query results to obtain a set of partitioned words, and determining the second set of query keywords based on selection frequencies of the partitioned words.
  • a query is made using at least some of the second set of query keywords, and query results are obtained.
  • the query keywords are input into a search engine to obtain the query results.
  • query results based on the second set of query keywords are used to update the user feedback database from which the user feedback log is obtained.
  • the first set of query keywords belong to one or more of the following types:
  • Type 1 words adjacent to the partitioned words.
  • Pl the statistical result related to the selection frequency in such a case.
  • Type 2 words including a portion of the partitioned words.
  • the statistical result related to the selection frequency in such a case is denoted as P2 in the following description.
  • Type 3 words including partitioned word.
  • the statistical result related to the selection frequency in such a case is denoted as P3 in the following description.
  • the user feedback log is adapted to record information corresponding to the query word such as historical query results and click frequencies, and display frequencies of the historical query results; for example, the query result may be a web page.
  • the query result may be a web page.
  • a resulting web page with higher selection frequency is deemed to be more relevant to the query word.
  • the latent word meaning of a word refers to another word that is synonymous, near-synonymous or partially synonymous to the original word, for example, " M#S" and "f ⁇ ffi ⁇ " ("glass bottle” and “glass bottles”); in another example, " ⁇ A*",
  • the first type of latent word meanings are words that often appear in pairs, for example, "Motorola” and “company” ("Slft ⁇ S” and “ ⁇ W]”), "Motorola” and “mobile phone” (" ⁇ ft ⁇ fe” and " ⁇ #1"). Such a relationship indicates a correlation between two words. Stated another way, some words after word partitioning are deemed to be adjacent (at least in meaning) to the query word.
  • the second type of latent word meaning for a word includes a plurality of words that appear in a certain order.
  • the query keyword "glass bottle” has corresponding latent words “glass” and “bottle”
  • "S? ⁇ IS” has latent meaning words "?£i! ⁇ ” and “M ⁇ ?”
  • "Jt&” has corresponding latent words "MBB lt ⁇ ” (beautiful), "& A” ( woman), that is, the partitioned words includes at least a portion of the query word after word partitioning.
  • the third type of latent word meaning is a word or character that forms another word. For example, "#F” (shrimp) and "Xj ⁇ F"
  • Such latent word meanings automatically found via user feedback such as click frequency usually represents the intention of the searched key word input by a user, which can be used to improve the accuracy rate of the search engine.
  • "bed"
  • the actual intention of the user is usually a bed for sleeping, for example, "#- A ⁇ " ("single bed”), “M A ⁇ ” ("double bed”), and "*t ⁇ ” ("wooden bed”), rather than a general support structure such as "?£ ⁇ " ("flower bed”) or some kind of machinery (“lathe”).
  • the first set of query keywords is obtained according to steps 102 and 104 and the historical query results (web page URLs, file IDs, etc.), the click frequencies, and display frequencies of the historical query results or a combination thereof is obtained according to step 106.
  • Word partitioning is performed on the first set of query keywords. If the first set of query keywords includes a plurality of words, the corresponding historical query results and related information in the user feedback log of the query words are added to each user feedback log corresponding to each partitioned word. Thus, each partitioned word will have its own corresponding historical query results. As a result of the processing, each query of the user feedback log will be an independent partitioned word.
  • the above processing related to Pl, P2, P3 will be respectively performed on each partitioned word or a part of the partitioned word and the selection of historical query results may be determined according to the information of total query times, click times, exposure times of the historical query results, or a combination thereof.
  • the historical query results corresponding to the partitioned words are processed respectively. All character strings that totally match the partitioned words are found in the historical query results in the user feedback log (here, totally matching means that the partitioned word is a substring of the character string) and the length of the character string may be the length of a sentence that contains the partitioned words, or M times of the length of a sentence that contains the partitioned words, wherein M may be an arbitrary integer greater than 1.
  • the corresponding statistical result of Pl, P2, or P3 is computed for each partitioned word.
  • files are used as the query results in the following embodiments, although other results such as URLs or other appropriate search objects can be used as well.
  • the selection frequencies of the query results and/or the selection frequencies of the contents of the query results may be considered.
  • the first set of query keywords is obtained based on user input according to steps 102 and 104 and historical query results (web page URLs and file IDs, etc.), click frequencies, display frequencies of the historical query results, or a combination thereof is obtained from user feedback log according to step 106.
  • a query dictionary may be optionally preconfigured and information corresponding to different search words, such as historical query results (web page URLs and file IDs, etc.), click frequencies, display frequencies of the historical query results, or a combination thereof may be input in advance.
  • the second set of query keywords may be obtained quickly and easily using the query dictionary. That is, the content of the previous user feedback log is pre-stored for query and the query dictionary may also be updated according to a new user feedback log; however, the user feedback log may also be invoked after the first query key word is input.
  • Type 1 partitioned words being adjacent to the first set of query keywords. If the first set of query keywords includes a partitioned word of a character string, for example, the first query keyword is " jlic" and the historical query result in the user feedback log is a file that includes the content " ⁇ ffl
  • Type 2 partitioned words include portions of words in the first set of query keywords. If the first set of query keywords is included in a plurality of adjacent word partitioning results. For example, the first query keyword is "H ⁇ T" (beauty) and the historical query result in the user feedback log has the partitioned content " HtM ⁇ l ;H:
  • the number of appearances of a partitioned word that includes the first query keyword appears is multiplied by the click frequency and/or display frequency of the query result, the product is taken as a coefficient of weight, marked as times weight (2), and it is added to the count P2 of the total query results.
  • P2 includes the weight (2) of the number of times a plurality of partitioned words of the first set of query keywords appears in the same order. Therefore, in this example, if the weight of the file is 0.3, the result corresponding to "HM$J
  • Type 3 partitioned words include the first set of query keywords. If the first query keyword is a substring of a partitioned word, for example, the query word is "#Kfi" (fridge) and the historical query result in the user feedback log has the partitioned content "
  • the number of appearances of the partitioned word containing the first query keyword is multiplied by the click frequency and/or display frequency of the query result, the product is taken as a coefficient of weight, marked as times weight (3), and it is added to count P3 of the total query result, wherein P3 is the weight (3) of the number of appearances of a partitioned word including the first query keyword in the query result. Therefore, in this example, if the weight of the file is 0.8, the result corresponding to " ⁇ ,/ ⁇ k$I" (refrigerator) in P3 is increased by 0.8.
  • one or more of the latent word meanings are processed.
  • the first, second, and third thresholds may be fixed thresholds or dynamically adjusted according to the overall query results of the query word. For example, weighted summation is performed on all query result files that contain the matched character string and then the result obtained is multiplied by a coefficient, which may be dynamically set according to the query results.
  • a threshold By setting a threshold, the words of the latent word meaning of a part of the query word are selectively determined instead of being an unconditionally feedback.
  • the selection frequencies may be the selection frequencies of users on the historical query results and it may also be the selection frequencies of users on the contents of historical query results.
  • the click frequency and/or display frequency of a file or its content is taken as a coefficient of weight.
  • the coefficient may be one or both of the click frequency and the display frequency and the value of the coefficient may have a linear or nonlinear relation with the click and display frequency.
  • the coefficient wherein both frequencies are greater than a given threshold is set to 1, otherwise, the coefficient is 0; or the coefficient with the highest click frequency and display frequency is 1 and other coefficients are normalized to [0,1] by being divided by the maximum value.
  • the purpose for selecting the selection frequency is to find the latent word meaning using the selection frequency. Some information with low selection frequency may be filtered by the given threshold, thereby improving the speed of the finding of the latent word meaning and avoiding the interference of some information at the same time.
  • the selection frequencies include one or more of: click frequencies on the historical query results, display frequencies on the historical query results, reading time on the historical query results, and importance of the historical query results.
  • the click frequency and display frequency (or one of them) of the file is taken as a coefficient of weight and the coefficient may also be other information of the file, for example, a combination of the amount of time users spent reading the file, a degree of importance, the click frequency, and the display frequency.
  • relationship between the query word and its latent word meaning is commutative.
  • #£-?”" glass
  • Different techniques may be used to determine the first set of query keywords in step 104.
  • One technique involves word partitioning being first performed on the information content that is input by a user and the partitioned words are used to obtain the query keywords.
  • the information content is decomposed into characters and each character is used as a query keyword.
  • Different ways of keyword determination may be performed at the same time, or combined.
  • word partitioning is first performed on the query word input by a user.
  • Word partitioning result query is performed using the partitioned words.
  • the latent semantics of the partitioned words of the query words are used to perform additional query and finally separate word query may be performed.
  • the correlated result is queried via "query word + word of the first latent word meaning", if the query word is " ⁇ ft ⁇ U" (Motorola), the corresponding word of the first latent word meaning is "Hft ⁇ JS ⁇ n]"
  • latent word meanings of "Uft ⁇ £&” (Motorola) are " y ⁇ W (company) and "##L" (mobile phone).
  • the query result is obtained using "adjacent query word” of the second latent word meaning, for example, the corresponding words of the second latent word meaning of "32£ ⁇ f ⁇ #&” (glass) are "Jj& ⁇ fii " (glass I bottles).
  • the query result is obtained via word of the third latent word meaning, for example, when " ⁇ , /?K $1 "
  • the correlation degree between a query word and a file is calculated based on the query result of latent word meaning query, it should be higher than that of the result obtained via separate word query.
  • the value of the correlation degree will influence the ordering of the query result (according to the correlation degree and web page importance degree, etc.).
  • the user characteristics of the user that inputs the first set of query keywords may also be obtained; that is, the user characteristics of the user may be obtained when the user inputs the first set of query keywords.
  • the user feedback log when the user feedback log is obtained in step 106, the user feedback log may also be obtained according to the user characteristics.
  • the user feedback log obtained includes historical query results that take words in the first set of query keywords as the target and the selection frequencies of users on the historical query results, while these historical query results include the user characteristic.
  • the second set of query keywords may be determined according to the user characteristic.
  • a different second set of query keywords may also be determined according to the user characteristic of the user that inputs the first set of query keywords.
  • the user characteristic By selecting the user feedback log using the user characteristic, it is favorable to further find the latent word meaning of the first set of query keywords. For example, according to the above embodiments, when a user searches for "J ⁇ ", the actual intention of the user is mostly a bed for sleeping, such as "#- ⁇ " (single bed), "3XAJ ⁇ ” (double bed) and " ⁇ +£1 ⁇ ” (wooden
  • the first set of query keywords that is input is " ⁇ jfc"
  • the user characteristic may include user region (for example, country, area and, town), previous web pages browsed by the user frequently, web pages browsed by the user lately, previous search keywords input by the user, user gender, age, career, and taste, etc.
  • technical measures such as IP address analysis, user-side browser historical data analysis, user-side COOKIE data analysis and user web registration information analysis, etc., may be used as required and this may be easily understood by one skilled in the art.
  • this application further provides a method and a device for word matching and a device for information query. Because the method and device for word matching, the device for information query and the method for information query are based on the same inventive conception, they have similar principles. Therefore, in the implementation of the method and device for word matching and the device for information query, reference may be made to the implementation of the method for information query and the repeated parts will not be described again here. [0052] The implementation of an embodiment of a query processing system is described below. The system is described to include a number of modules, which in turn includes submodules or units.
  • the modules/submodules/units can be implemented as software components executing on one or more processors, as hardware such as programmable logic devices and/or Application Specific Integrated Circuits designed to perform certain functions or a combination thereof.
  • the modules can be embodied by a form of software products which can be stored in a nonvolatile storage medium (such as optical disk, flash storage device, mobile hard disk, etc.), including a number of instructions for making a computer device (such as personal computers, servers, network equipments, etc.) implement the methods described in the embodiments of the present invention.
  • the modules/submodules/units may be implemented on a single device or distributed across multiple devices. The functions of the modules may be merged into one another or further split into multiple sub-modules.
  • FIG. 2 is a block diagram illustrating an embodiment of a query processing system.
  • the system includes: a first-query-keyword obtaining module 201, for obtaining the first set of query keywords based on user input; a user- feedback-log obtaining module 202, for obtaining a user feedback log of the first set of query keywords; a matching module 203, for determining the second set of query keywords that matches the first set of query keywords according to the user feedback log; and a query- result feedback module 204, for feeding back a query result that takes the second set of query keywords as a target.
  • a first-query-keyword obtaining module 201 for obtaining the first set of query keywords based on user input
  • a user- feedback-log obtaining module 202 for obtaining a user feedback log of the first set of query keywords
  • a matching module 203 for determining the second set of query keywords that matches the first set of query keywords according to the user feedback log
  • a query- result feedback module 204 for feeding back a query result
  • the user-feedback-log obtaining module may be further adapted to obtain a user feedback log, which includes all previous historical query results that take words in the first set of query keywords as search targets and all previous selection frequencies of the user on the historical query results.
  • the matching module may be further adapted to determine the second set of query keywords that matches the first set of query keywords according to the historical query results and selection frequencies in the user feedback log.
  • the user-feedback-log obtaining module may be further adapted to obtain, as the selection frequencies, the selection frequencies on historical query results and/or the selection frequencies on the contents of historical query results.
  • FIG. 3 is a block diagram illustrating an embodiment of the matching module.
  • the matching module includes: a content obtaining unit 2031, for obtaining the contents of historical query results corresponding to the first set of keywords; a word partitioning unit 2032, for carrying out word partitioning on the contents of historical query results, and obtaining words after word partitioning; and a matching unit 2033, for determining the second set of query keywords that matches the first set of query keywords according to the selection frequencies of the words after word partitioning.
  • the word partitioning unit may also be further adapted to obtain, after word partitioning, words of one or more of the following types: words adjacent to the first set of query keywords after word partitioning; words containing the first set of query keywords after word partitioning; and words containing a component part of the first set of query keywords after word partitioning.
  • the matching module may be further adapted to determine the second set of query keywords that matches the first set of query keywords according to the historical query results and selection frequencies, wherein the selection frequencies are larger than a given threshold.
  • FIG. 4 is a block diagram illustrating an embodiment of the first-query- keyword obtaining module.
  • the first-query-keyword obtaining module may include: an information-content obtaining unit 201 1, for obtaining the information content that is input by a user; a word partitioning/decomposing unit 2012, for obtaining words after word partitioning after carrying out word partitioning on the information content, and/or decomposing the information content into characters; and a first- query-keyword determining unit 2013, for taking the words and/or characters after word partitioning as the first set of query keywords.
  • the user-feedback-log obtaining module may be further adapted to obtain, as the selection frequencies, one or a combination of: click frequencies on the historical query results, display frequencies on the historical query results, reading time on the historical query results, and importance of the historical query results.
  • the first-query-keyword obtaining module may be further adapted to obtain the user characteristic of a user when the user inputs the first set of query keywords; the user-feedback-log obtaining module may be further adapted to obtain a user feedback log according to the user characteristic.
  • the first-query-keyword obtaining module may be further adapted to obtain the user characteristic of a user when the user inputs the first set of query keywords;
  • the user-feedback-log obtaining module may also be further adapted to, when obtaining a user feedback log, obtain a user feedback log that includes historical query results that take the word to be matched as an object and the selection frequencies of the user on the historical query results, wherein the historical query results include the user characteristic.
  • the first-query-keyword obtaining module may also be further adapted to obtain the user characteristic of a user when the user inputs the first set of query keywords.
  • the matching module may be further adapted to determine the second set of query keywords according to the user characteristic when determining the second set of query keywords according to the user feedback log.
  • FIG.5 is a flowchart illustrating an embodiment of a process for word matching. As shown in the figure, it may include the following steps when word matching is performed: at 501, obtaining a word to be matched, such as the first set of query keywords; at 502, obtaining a user feedback log according to the word to be matched, wherein the user feedback log includes all previous historical query results that take the word to be matched as an object and all previous selection frequencies of users on the historical query results; at 503, determining words that match the word to be matched according to the historical query results and selection frequencies.
  • word matching is performed: at 501, obtaining a word to be matched, such as the first set of query keywords; at 502, obtaining a user feedback log according to the word to be matched, wherein the user feedback log includes all previous historical query results that take the word to be matched as an object and all previous selection frequencies of users on the historical query results; at 503, determining words that match the word to be matched according to the historical query results and selection frequencies.
  • the selection frequencies include: selection frequencies on historical query results and/or selection frequencies on the contents of historical query results.
  • the process of determining words that match the word to be matched according to the selection frequencies on the contents of historical query results includes the following: obtaining the contents of historical query results of the word to be matched; carrying out word partitioning on the contents of historical query results, and obtaining words after word partitioning; and determining words that match the word to be matched according to the selection frequencies of the words after word partitioning.
  • words after word partitioning refer to words of one or more of the following types: words adjacent to the word to be matched after word partitioning; words containing the word to be matched after word partitioning; and words containing a component part of the word to be matched after word partitioning.
  • the selection frequencies are larger than a given threshold.
  • the process of obtaining a keyword to be matched may be as follows: obtaining the information content that is input by a user; obtaining words after word partitioning after carrying out word partitioning on the information content and/or decomposing the information content into characters; and taking the words and/or characters after word partitioning as the word to be matched.
  • the selection frequencies may include one or a combination of the following: click frequencies on the historical query results, display frequencies on the historical query results, reading time on the historical query results, and importance of the historical query results.
  • the method may also further include the steps of obtaining the user characteristic of a user when the user inputs a word to be matched and obtaining a user feedback log according to the user characteristic when obtaining the user feedback log.
  • the method may also further include the following steps: obtaining the user characteristic of a user when the user inputs a word to be matched; obtaining a user feedback log, wherein the user feedback log obtained includes historical query results that take the word to be matched as an object and the selection frequencies of users on the historical query results; and the historical query results include the user characteristic.
  • the method may also further include the steps of obtaining the user characteristic of a user when the user inputs a word to be matched and determining words that match the word to be matched according to the user characteristic when determining words that match the word to be matched according to the user feedback log.
  • FIG.6 is a block diagram illustrating an embodiment of a device for word matching.
  • the device includes: a word-to-be-matched obtaining module 601, for obtaining a word to be matched; a user- feedback-log obtaining module 602, for obtaining a user feedback log according to the word to be matched; and a matching module 603, for determining words that match the word to be matched according to the user feedback log.
  • the user-feedback-log obtaining module may be further adapted to obtain a user feedback log, which includes all previous historical query results that take the word to be matched as a target and all previous selection frequencies of users on the historical query results; and the matching module may be further adapted to determine words that match the word to be matched according to the historical query results and selection frequency in the user feedback log.
  • the user-feedback-log obtaining module may be further adapted to obtain, as the selection frequencies, the selection frequencies on historical query results and/or the selection frequencies on the contents of historical query results.
  • the matching module includes: a content obtaining unit, for obtaining the contents of historical query results of the word to be matched; a word partitioning unit, for carrying out word partitioning on the contents of historical query results and obtaining words after word partitioning; and a matching unit, for determining words that match the word to be matched according to the selection frequencies of the words after word partitioning.
  • the word partitioning unit may be further adapted to obtain, after word partitioning, words in the following modes or a combination thereof: words adjacent to the word to be matched after word partitioning; words containing the word to be matched after word partitioning; and words containing a component part of the word to be matched after word partitioning.
  • the matching module may be further adapted to determine words that match the word to be matched according to the historical query results and selection frequencies, wherein the selection frequencies are larger than a given threshold.
  • the word-to-be-matched obtaining module may include: an information- content obtaining unit, for obtaining the information content that is input by a user; a word partitioning/decomposing unit, for obtaining words after word partitioning after carrying out word partitioning on the information content and/or decomposing the information content into characters; and a word-to-be-matched determining unit, for taking the words and/or characters after word partitioning as the word to be matched.
  • the user- feedback-log obtaining module may be further adapted to obtain, as the selection frequencies, one or a combination of: click frequencies on the historical query results, display frequencies on the historical query results, reading time on the historical query results, and importance of the historical query results.
  • the word-to-be-matched obtaining module is further adapted to obtain the user characteristic of a user when the user inputs a word to be matched; the user- feedback-log obtaining module is further adapted to obtain a user feedback log according to the user characteristic.
  • the word-to-be-matched obtaining module may be further adapted to obtain the user characteristic of a user when the user inputs a word to be matched and the user-feedback-log obtaining module may also be further adapted to, when obtaining a user feedback log, obtain a user feedback log that includes historical query results that take the word to be matched as an object and the selection frequencies of users on the historical query results, wherein the historical query results include the user characteristic.
  • the word-to-be-matched obtaining module may also be further adapted to obtain the user characteristic of a user when the user inputs a word to be matched and the matching module may be further adapted to, when determining words that match the word to be matched according to the user feedback log, determine words that match the word to be matched according to the user characteristic.
  • the latent semantics of a word may be automatically found and the hidden relation between words can be found correctly. Further, the latent semantics of a word automatically found and the correlated semantics of a word to be queried may be used to improve the effect of the search engine. Moreover, when the latent word meaning of a query word is automatically found, a similar effect may also be obtained according to the word frequencies of separate words immediately adjacent to the query word in the search string, rather than only according to the word partitioning result. Therefore, in an embodiment of this application, the performance of the search engine may be improved via the latent word meaning of a word that is automatically found. In comparison with the traditional techniques, search accuracy and efficiency may be improved.
  • this application may be provided as a method, a system, or a computer program product. Therefore, this application may be in the form of full hardware embodiments, full software embodiments, or a combination thereof. Moreover, this application may be in the form of a computer program product that is implemented on one or more computer-usable storage media (including, without limitation, magnetic disk storage, CD-ROM and optical storage) containing computer-usable program codes.
  • computer-usable storage media including, without limitation, magnetic disk storage, CD-ROM and optical storage
  • each flow and/or block in the flow chart and/or block diagram and the combination of flow and/or block in the flow chart and/or block diagram may be realized via computer program instructions.
  • Such computer program instructions may be provided to the processor of a general-purpose computer, special-purpose computer, a built-in processor or other programmable data processing devices, to produce a machine, so that the instructions executed by the processor of a computer or other programmable data processing devices may produce a device for realizing the functions specified in one or more flows in the flow chart and/or one or more blocks in the block diagram.
  • Such computer program instructions may also be stored in a computer- readable storage that can guide a computer or other programmable data processing devices to work in a specific mode, so that the instructions stored in the computer-readable storage may produce a manufacture including a commander equipment, wherein the commander equipment may realize the functions specified in one or more flows of the flow chart and one or more blocks in the block diagram.
  • Such computer program instructions may also be loaded to a computer or other programmable data processing devices, so that a series of operational processes may be executed on the computer or other programmable devices to produce a computer-realized processing, thereby the instructions executed on the computer or other programmable devices may provide a process for realizing the functions specified in one or more flows in the flow chart and/or one or more blocks in the block diagram.

Abstract

Performing a query includes receiving a user input, determining a first set of query keywords based at least in part on the user input, obtaining, based on at least some of the first set of query keywords, a user feedback log that includes historical query results, determining a second set of query keywords based on the user feedback log, and making a query based on at least some of the second set of query keywords.

Description

SEARCH ENGINE FOR REFINING CONTEXT-BASED QUERIES BASED UPON HISTORICAL USER FEEDBACK
[00011 This application claims priority to People's Republic of China Patent
Application 200910001555.7 entitled METHOD AND SYSTEM FOR WORD MATCHING AND INFORMATION QUERY, filed January 12, 2009 which is incorporated herein by reference for all purposes.
FIELD OF THE INVENTION
[0002) The present application relates to data processing technology and, in particular, to a method and a device for word matching and information query.
BACKGROUND OF THE INVENTION
[0003] Latent word meaning usually refers to the latent meaning of a word or phrase and it usually may be expressed in one or more words or phrases. For example, the latent word meaning of "#KH" (fridge) generally refers to
Figure imgf000002_0001
(refrigerator), while the latent word meaning of "f&tfø " (cotton slippers) generally refers to " ^t^ tfe ^" (all-cotton slippers), etc.
[0004] Many studies have been carried out on the automatic finding of latent semantics, most of which attempt to find near-synonym using the co-appearance and link relation of words. Some existing techniques use synonyms to determine the relationship between words. The number of vocabulary entries obtained based on manual-labeled corpora, however, is limited and it can be difficult to guarantee the effect of automatic finding of synonyms.
[0005] The indexing mode of a search engine typically includes separate word search, word partitioning indexing, and hybrid indexing. When using the separate word indexing technique, the distance between the separate words in a file typically needs to be calculated. Thus, the efficiency is often low and accuracy often poor. The problem is particularly pronounced in languages without natural word separators (e.g., spaces) between words, such as Chinese. For example, the differences among "&!§" (pesticides), "}Φ3?iξ P" (Shen Nong pharmaceuticals) and "#&&i§J~" (Shen Nong pesticides factory) cannot be readily distinguished using separate word indexing. In contrast, the word partitioning search technique has higher accuracy and is fast, but the recall rate is often low. For example, when it searches for "#K$f " (fridge), only the results for "#Kϋ" (fridge) can be found and the results of "%/7K$t" (refrigerator) cannot be found. For the hybrid indexing method in which the separate word indexing and the word partitioning indexing are combined, it usually first queries according to the word partitioning indexing technique and then queries according to the separate word indexing technique. For example, when "^^ #5" (glass bottle) is queried, the results of "l£ϊfij}fδ" (glass bottle) are found according to word partitioning indexing and then other results are found according to separate word indexing. This makes up for the disadvantages of the two earlier-mentioned methods, but " (glass bottles) is found according to separate word indexing and the search engine cannot distinguish between
Figure imgf000003_0001
(causing bottle neck), therefore the accuracy is affected. More effective search techniques are therefore needed.
BRIEF DESCRIPTION OF THE DRAWINGS
[0006] Various embodiments of the invention are disclosed in the following detailed description and the accompanying drawings.
[0007] FIG. IA is a block diagram illustrating an embodiment of a query processing system.
[0008] FIG. IB is a flowchart of an embodiment of a process for performing a query.
[0009] FIG. 2 is a block diagram illustrating an embodiment of a query processing system.
[0010] FIG. 3 is a block diagram illustrating an embodiment of the matching module.
[0011| FIG. 4 is a block diagram illustrating an embodiment of the first-query- keyword obtaining module.
[0012] FIG.5 is a flowchart illustrating an embodiment of a process for word matching.
[0013] FIG.6 is a block diagram illustrating an embodiment of a device for word matching.
DETAILED DESCRIPTION
[0014] The invention can be implemented in numerous ways, including as a process; an apparatus; a system; a composition of matter; a computer program product embodied on a computer readable storage medium; and/or a processor, such as a processor configured to execute instructions stored on and/or provided by a memory coupled to the processor. In this specification, these implementations, or any other form that the invention may take, may be referred to as techniques. In general, the order of the steps of disclosed processes may be altered within the scope of the invention. Unless stated otherwise, a component such as a processor or a memory described as being configured to perform a task may be implemented as a general component that is temporarily configured to perform the task at a given time or a specific component that is manufactured to perform the task. As used herein, the term 'processor' refers to one or more devices, circuits, and/or processing cores configured to process data, such as computer program instructions.
[0015] A detailed description of one or more embodiments of the invention is provided below along with accompanying figures that illustrate the principles of the invention. The invention is described in connection with such embodiments, but the invention is not limited to any embodiment. The scope of the invention is limited only by the claims and the invention encompasses numerous alternatives, modifications and equivalents. Numerous specific details are set forth in the following description in order to provide a thorough understanding of the invention. These details are provided for the purpose of example and the invention may be practiced according to the claims without some or all of these specific details. For the purpose of clarity, technical material that is known in the technical fields related to the invention has not been described in detail so that the invention is not unnecessarily obscured.
[0016] FIG. IA is a block diagram illustrating an embodiment of a query processing system. In the example shown, query processing system 150 includes a query processing server 152, which in some embodiments include one or more devices having one or more processors coupled to one or more memories. Query processing server 152 includes one or more interfaces configured to receive user inputs, for example user inputs sent by client devices such as 154 over a network 155, which may be the Internet or other communications network. Examples of a communication interface include without limitation external connections, such as a port, cable, wired or wireless network interface card, etc., and internal connections, such as a communication bus. In some embodiments, the client device is a computing device such as a computer or mobile device with networking capabilities. Query processing server 152 has access to a user feedback database 156 that stores user feedback logs.
[0017] As will be described in greater detail below, upon receiving user input, the query processing server processes the user input to generate query keywords based on the user input and user feedback log. In some embodiments, the query keywords are sent to a search engine 158 which locates and returns search results based on the query keywords. In some embodiments, the search engine is configured to search for specific content stored in a database, such as product information on an electronic commerce website. In some embodiments, the search engine is configured as a general purpose search engine that searches for content on the Internet. The functions of the keyword search engine and the query processing server are combined in some embodiments; thus, the query processing server uses the keywords to perform search and returns the results directly. The search results and user response to the search results (such as user selection/click-through) are updated in the user feedback database.
[0018] FIG. IB is a flowchart of an embodiment of a process for performing a query.
Process 100 may be performed on a system such as 150. At 102, a user input that includes a plurality of symbols is received. In some embodiments, the user input is sent from a client device such as 154 via a web browser or other user interface tool to the query processing server. As used herein, a symbol refers to a character or a letter. A number of characters/letters form a word.
[0019] At 104, a first set of query keywords are determined based on the user input.
The first set of query keywords may include one or more words. In some embodiments, the user input is used directly as the first set of query keywords. In some embodiments, the user input is partitioned and the partitioned words are used as the first set of query keywords. In some cases the user input is in a letter based language such as English, where words are naturally separated by spaces. The user input therefore does not require context based processing and the first set of query keywords include the words in the user input and/or stemmed input words. In some cases, character based languages such as Chinese is used as the user input language. Since there are no specific character dividers in such languages, the input symbols are partitioned into words based on context, and the first set of query keywords include the partitioned words. In some embodiments, the first set of query keywords also includes the characters in the user input, which are available without any partitioning.
[0020] At 106, a user feedback log is obtained based on the first set of query keywords. The user feedback log, which is obtained from a database such as 156 in some embodiments, includes historical query results using the first set of query keywords as the target (or object) of the query and the selection frequency (also referred to as the click- through rate) by the users. In various embodiments, the selection frequency may include selection frequency on the previous query results and/or selection frequency on the contents of previous results. The user feedback log is gathered by the search engine. In some embodiments, keywords input by the users, past query results (e.g., web page file identifiers/universal resource locators (URLs)), click frequencies on the past query results, display frequencies of the past query results, etc. are collected by the search engine and stored. Suppose, for example, the first set of query keywords include the words "western medicine" or "MI5", then the user feedback log includes records of all or some of the results
returned when users previously made queries using the words "western medicine" or "MI5".
[0021] At 108, a second set of query keywords that matches the first set of query keywords based on the user feedback log is determined. In some embodiments, the second set of query keywords includes latent meaning words derived based on the user feedback log. All or a portion (such as a randomly selected portion) of the previous records may be used to establish latent word meaning. The automatic finding of latent words with latent meaning refers to the finding of one or more words (or phrases) with relevant or similar word meaning as the original word. Based on user feedback logs, the latent word meaning relationship between the query word and the historical query results that embody the user's intention can be reliably and automatically found. The latent word meaning relationship is in turn used to improve the accuracy rate and intelligence of the search engine. Therefore, the user feedback log may include all previous historical query results that use the first set of query keywords as the search target and all previous click selecting frequencies of users on the historical query results.
[0022] In some embodiments, determining the second set of query keywords includes: obtaining the content of historical query results of the first set of keywords, partitioning the content of the historical query results to obtain a set of partitioned words, and determining the second set of query keywords based on selection frequencies of the partitioned words.
(0023] At 110, a query is made using at least some of the second set of query keywords, and query results are obtained. In some embodiments, the query keywords are input into a search engine to obtain the query results.
[0024 j At 1 12, query results based on the second set of query keywords are used to update the user feedback database from which the user feedback log is obtained.
[0025] In various embodiments, the first set of query keywords belong to one or more of the following types:
[0026] Type 1 : words adjacent to the partitioned words. For purposes of illustration, the statistical result related to the selection frequency in such a case is denoted as Pl in the following description.
[0027] Type 2: words including a portion of the partitioned words. The statistical result related to the selection frequency in such a case is denoted as P2 in the following description.
[0028] Type 3: words including partitioned word. The statistical result related to the selection frequency in such a case is denoted as P3 in the following description.
[0029] The user feedback log is adapted to record information corresponding to the query word such as historical query results and click frequencies, and display frequencies of the historical query results; for example, the query result may be a web page. For a given query word, a resulting web page with higher selection frequency is deemed to be more relevant to the query word. The latent word meaning of a word refers to another word that is synonymous, near-synonymous or partially synonymous to the original word, for example, " M#S" and "fø^ffi÷" ("glass bottle" and "glass bottles"); in another example, "^A*",
"-Φ-AΛ" and "ϋϋøξ" ("double bed", "single bed", "mattress bed") all have the latent word meaning of "J^" ("bed") which is a place for people to sleep, but "?£^" ("flower bed") does not have the latent word meanings of "$?" ("bed").
[0030] At least three types of latent word meanings are used in various embodiments of this application. The first type of latent word meanings are words that often appear in pairs, for example, "Motorola" and "company" ("Slft^S" and "^W]"), "Motorola" and "mobile phone" ("øft^ fe" and "^#1"). Such a relationship indicates a correlation between two words. Stated another way, some words after word partitioning are deemed to be adjacent (at least in meaning) to the query word. The second type of latent word meaning for a word includes a plurality of words that appear in a certain order. For example, the query keyword "glass bottle" has corresponding latent words "glass" and "bottle", "S?^IS" has latent meaning words "?£i!§" and "M ?", and "Jt&" (beauty) has corresponding latent words "MBB ltø" (beautiful), "& A" (woman), that is, the partitioned words includes at least a portion of the query word after word partitioning. The third type of latent word meaning is a word or character that forms another word. For example, "#F" (shrimp) and "Xj ΪF"
(prawns), "J@" (alcohol) and "H$)@" (beer), that is, the partitioned words include the query keyword. Such latent word meanings automatically found via user feedback such as click frequency usually represents the intention of the searched key word input by a user, which can be used to improve the accuracy rate of the search engine. For example, when a user searches for "^" ("bed"), the actual intention of the user is usually a bed for sleeping, for example, "#- A^" ("single bed"), "M A^" ("double bed"), and "*tø^" ("wooden bed"), rather than a general support structure such as "?£^" ("flower bed") or some kind of machinery ("lathe"). Based on user feedback such as selection of displayed search results, it can be determined that that "#- Aflξ" ("single bed"), "M Aft" ("double bed"), and "?fc$iΛ" ("wooden bed") have a latent word meaning of "J^" (bed for sleeping), while as " ?£^" ("flower bed") does not has a latent word meaning of "J^".
[0031] In some embodiments, the first set of query keywords is obtained according to steps 102 and 104 and the historical query results (web page URLs, file IDs, etc.), the click frequencies, and display frequencies of the historical query results or a combination thereof is obtained according to step 106. Word partitioning is performed on the first set of query keywords. If the first set of query keywords includes a plurality of words, the corresponding historical query results and related information in the user feedback log of the query words are added to each user feedback log corresponding to each partitioned word. Thus, each partitioned word will have its own corresponding historical query results. As a result of the processing, each query of the user feedback log will be an independent partitioned word. The above processing related to Pl, P2, P3 will be respectively performed on each partitioned word or a part of the partitioned word and the selection of historical query results may be determined according to the information of total query times, click times, exposure times of the historical query results, or a combination thereof. The historical query results corresponding to the partitioned words are processed respectively. All character strings that totally match the partitioned words are found in the historical query results in the user feedback log (here, totally matching means that the partitioned word is a substring of the character string) and the length of the character string may be the length of a sentence that contains the partitioned words, or M times of the length of a sentence that contains the partitioned words, wherein M may be an arbitrary integer greater than 1. Depending on the type of the partitioned word, the corresponding statistical result of Pl, P2, or P3 is computed for each partitioned word. For purposes of illustration, files are used as the query results in the following embodiments, although other results such as URLs or other appropriate search objects can be used as well. During implementation, the selection frequencies of the query results and/or the selection frequencies of the contents of the query results may be considered.
[0032] In some embodiments, the first set of query keywords is obtained based on user input according to steps 102 and 104 and historical query results (web page URLs and file IDs, etc.), click frequencies, display frequencies of the historical query results, or a combination thereof is obtained from user feedback log according to step 106. Meanwhile, a query dictionary may be optionally preconfigured and information corresponding to different search words, such as historical query results (web page URLs and file IDs, etc.), click frequencies, display frequencies of the historical query results, or a combination thereof may be input in advance. This way, when the first set of query keywords is input, the second set of query keywords may be obtained quickly and easily using the query dictionary. That is, the content of the previous user feedback log is pre-stored for query and the query dictionary may also be updated according to a new user feedback log; however, the user feedback log may also be invoked after the first query key word is input.
[0033] The processing of different types of partitioned words is discussed.
[0034] Type 1 : partitioned words being adjacent to the first set of query keywords. If the first set of query keywords includes a partitioned word of a character string, for example, the first query keyword is " jlic" and the historical query result in the user feedback log is a file that includes the content "Φ ffl|*ft|U£:|ffiJU|£ |^3te|, |##C|tøil |tf Hfl|ttl£"
(Chinese | ancient | beauty | Xishi | named | Yiguan |, born | during | Chunqiu | Zhanguo | period" (here, "|" represents the place where word partitioning occurs). Here, the number of appearances (T) of the partitioned words immediately adjacent to the query keyword in the query result string is multiplied by the click frequency and/or display frequency of the query result, and the product is taken as a coefficient of weight, marked as times weight (1), and is added to the count Pl of the total query result. Pl includes the weight (1) of the number of times each word that appears in front of or behind the first query keyword. For example, in this example, if the weight of the file is 0.5, the result corresponding to "l*f ft" (ancient) and "HiS" (Xishi) in Pl (here, two words are immediately adjacent to the query keyword and T=I) will be correspondingly increased by 0.5.
[0035] Type 2: partitioned words include portions of words in the first set of query keywords. If the first set of query keywords is included in a plurality of adjacent word partitioning results. For example, the first query keyword is "H^T" (beauty) and the historical query result in the user feedback log has the partitioned content " HtMϊl ;H: |^| H M β^)|^cA|" (Xishi I is I a I beautiful | woman |). At this point, the number of appearances of a partitioned word that includes the first query keyword appears is multiplied by the click frequency and/or display frequency of the query result, the product is taken as a coefficient of weight, marked as times weight (2), and it is added to the count P2 of the total query results. P2 includes the weight (2) of the number of times a plurality of partitioned words of the first set of query keywords appears in the same order. Therefore, in this example, if the weight of the file is 0.3, the result corresponding to "HM$J|icΛ" in P2 is increased by 0.3.
[0036] Type 3: partitioned words include the first set of query keywords. If the first query keyword is a substring of a partitioned word, for example, the query word is "#Kfi" (fridge) and the historical query result in the user feedback log has the partitioned content "
^L^^II^-fi^lUII^Sll-^H^ii^" (refrigerator | air conditioning | maintenance | and | repair).
At this point, the number of appearances of the partitioned word containing the first query keyword is multiplied by the click frequency and/or display frequency of the query result, the product is taken as a coefficient of weight, marked as times weight (3), and it is added to count P3 of the total query result, wherein P3 is the weight (3) of the number of appearances of a partitioned word including the first query keyword in the query result. Therefore, in this example, if the weight of the file is 0.8, the result corresponding to "Φ,/τk$I" (refrigerator) in P3 is increased by 0.8.
[0037] The processing steps are repeated until every partitioned word in all the historical query results in the user feedback log is processed. Based on the weighted sum of the number of times the partitioned word appears according to Pl, partitioned words of which the times weight sum is larger than a first threshold are taken as the first latent word meaning relation of the query word. Similarly, according to the weighted sum of the number of times partitioned words appear in P2 and P3, partitioned words of which the times weight sums are larger than a second threshold and a third threshold respectively are taken as the second latent word meaning and the third latent word meaning relation of the word.
[0038] In various embodiments, one or more of the latent word meanings are processed. The first, second, and third thresholds may be fixed thresholds or dynamically adjusted according to the overall query results of the query word. For example, weighted summation is performed on all query result files that contain the matched character string and then the result obtained is multiplied by a coefficient, which may be dynamically set according to the query results. By setting a threshold, the words of the latent word meaning of a part of the query word are selectively determined instead of being an unconditionally feedback.
[0039] In some embodiments, when the second set of query keywords that matches the first set of query keywords is determined according to the historical query results and selection frequencies, it may require that the selection frequencies be larger than the given threshold, wherein, the selection frequencies may be the selection frequencies of users on the historical query results and it may also be the selection frequencies of users on the contents of historical query results. The click frequency and/or display frequency of a file or its content is taken as a coefficient of weight. The coefficient may be one or both of the click frequency and the display frequency and the value of the coefficient may have a linear or nonlinear relation with the click and display frequency. For example, the coefficient wherein both frequencies are greater than a given threshold is set to 1, otherwise, the coefficient is 0; or the coefficient with the highest click frequency and display frequency is 1 and other coefficients are normalized to [0,1] by being divided by the maximum value. The purpose for selecting the selection frequency is to find the latent word meaning using the selection frequency. Some information with low selection frequency may be filtered by the given threshold, thereby improving the speed of the finding of the latent word meaning and avoiding the interference of some information at the same time.
[0040] In various embodiments, the selection frequencies include one or more of: click frequencies on the historical query results, display frequencies on the historical query results, reading time on the historical query results, and importance of the historical query results. The click frequency and display frequency (or one of them) of the file is taken as a coefficient of weight and the coefficient may also be other information of the file, for example, a combination of the amount of time users spent reading the file, a degree of importance, the click frequency, and the display frequency.
[0041] In implementation, relationship between the query word and its latent word meaning is commutative. For example, "3feϊ$j|#£-?"" (glass | bottles) is the latent word meaning of "5&MM" (glass bottle), and "3ft^#K" (glass bottle) is the latent word meaning of "WR\MΦ' (glass I bottles); "%3?MS" (refrigerator) is the latent word meaning of "/*$!
" (fridge), and
Figure imgf000012_0001
(refrigerator).
[0042] Different techniques may be used to determine the first set of query keywords in step 104. One technique involves word partitioning being first performed on the information content that is input by a user and the partitioned words are used to obtain the query keywords. In another technique, the information content is decomposed into characters and each character is used as a query keyword. Different ways of keyword determination may be performed at the same time, or combined. In the combined technique, word partitioning is first performed on the query word input by a user. Word partitioning result query is performed using the partitioned words. The latent semantics of the partitioned words of the query words are used to perform additional query and finally separate word query may be performed. Word partitioning result query refers to that the correlated result is queried from the word partitioning index according to the word partitioning result of the query word; separate word query refers to that the result is queried from the separate word indexing; and latent semantics query refers to that the query result is obtained using the latent meaning of the query word.
[0043] For a word of the first latent word meaning, the correlated result is queried via "query word + word of the first latent word meaning", if the query word is "^ft^U" (Motorola), the corresponding word of the first latent word meaning is "Hft^JS^ n]"
(Motorola company) and "IS^ES? J-Z^ #1" (Motorola mobile phone); here, the first type of
latent word meanings of "Uft^ £&" (Motorola) are "yΛW (company) and "##L" (mobile phone). For word of the second type of latent word meaning, the query result is obtained using "adjacent query word" of the second latent word meaning, for example, the corresponding words of the second latent word meaning of "32£ϊf§#&" (glass) are "Jj&ϊfii
Figure imgf000013_0001
" (glass I bottles). For word of the third type of latent word meaning, the query result is obtained via word of the third latent word meaning, for example, when " Φ, /?K $1 "
(refrigerator) is queried, the word of the third latent word meaning is "#KH" (fridge).
[0044] When the correlation degree between a query word and a file is calculated based on the query result of latent word meaning query, it should be higher than that of the result obtained via separate word query. The value of the correlation degree will influence the ordering of the query result (according to the correlation degree and web page importance degree, etc.).
[0045] In some embodiments, when the first set of query keywords is obtained in step 104, the user characteristics of the user that inputs the first set of query keywords may also be obtained; that is, the user characteristics of the user may be obtained when the user inputs the first set of query keywords.
[0046] Thus, when the user feedback log is obtained in step 106, the user feedback log may also be obtained according to the user characteristics.
[0047] Or, when the user feedback log is obtained, the user feedback log obtained includes historical query results that take words in the first set of query keywords as the target and the selection frequencies of users on the historical query results, while these historical query results include the user characteristic.
[0048| Or, when it determines the second set of query keywords according to the user feedback log, the second set of query keywords may be determined according to the user characteristic.
[0049] That is, when the second set of query keywords is matched according to the user feedback log, a different second set of query keywords may also be determined according to the user characteristic of the user that inputs the first set of query keywords. By selecting the user feedback log using the user characteristic, it is favorable to further find the latent word meaning of the first set of query keywords. For example, according to the above embodiments, when a user searches for "J^", the actual intention of the user is mostly a bed for sleeping, such as "#-ΛΛ" (single bed), "3XAJ^" (double bed) and "^+£1^" (wooden
bed), rather than a mechanical equipment such as
Figure imgf000014_0001
or "¥^" (lathe). At this point, it may be inferred based on the feedback of the user, such as user selection of content pertaining to a bed for sleeping, that the former has a latent word meaning of "J^" (bed) and the latent word meaning does not contain "ITLJ^" (lathe) and so on. However, for the same query keyword "J^" (bed), if the user is a technician in the field of mechanical equipment, the latent word meaning should be "#L$c" (lathe), rather than "#-Λj^" (single bed), "Mλ.
Figure imgf000014_0002
(wooden bed), etc. In this embodiment, "technician in the field of mechanical equipment" will be the user characteristic, which is used to classify the user feedback logs for better finding the latent word meaning of the word.
[0050] In another example, the first set of query keywords that is input is "^ jfc"
(apple), if the user characteristic is computer worker, the second set of query keywords in the category of computer is used as the matching set; if the user characteristic is food industry worker or agricultural worker, the second keyword in the category of fruit will be used as the matching set. During specific implementation, the user characteristic may include user region (for example, country, area and, town), previous web pages browsed by the user frequently, web pages browsed by the user lately, previous search keywords input by the user, user gender, age, career, and taste, etc. For the analyzing and classifying of user characteristics, technical measures such as IP address analysis, user-side browser historical data analysis, user-side COOKIE data analysis and user web registration information analysis, etc., may be used as required and this may be easily understood by one skilled in the art.
[0051] Based on the same inventive conception, this application further provides a method and a device for word matching and a device for information query. Because the method and device for word matching, the device for information query and the method for information query are based on the same inventive conception, they have similar principles. Therefore, in the implementation of the method and device for word matching and the device for information query, reference may be made to the implementation of the method for information query and the repeated parts will not be described again here. [0052] The implementation of an embodiment of a query processing system is described below. The system is described to include a number of modules, which in turn includes submodules or units. The modules/submodules/units can be implemented as software components executing on one or more processors, as hardware such as programmable logic devices and/or Application Specific Integrated Circuits designed to perform certain functions or a combination thereof. In some embodiments, the modules can be embodied by a form of software products which can be stored in a nonvolatile storage medium (such as optical disk, flash storage device, mobile hard disk, etc.), including a number of instructions for making a computer device (such as personal computers, servers, network equipments, etc.) implement the methods described in the embodiments of the present invention. The modules/submodules/units may be implemented on a single device or distributed across multiple devices. The functions of the modules may be merged into one another or further split into multiple sub-modules.
[0053] FIG. 2 is a block diagram illustrating an embodiment of a query processing system. As shown in the figure, the system includes: a first-query-keyword obtaining module 201, for obtaining the first set of query keywords based on user input; a user- feedback-log obtaining module 202, for obtaining a user feedback log of the first set of query keywords; a matching module 203, for determining the second set of query keywords that matches the first set of query keywords according to the user feedback log; and a query- result feedback module 204, for feeding back a query result that takes the second set of query keywords as a target.
[0054] In implementation, the user-feedback-log obtaining module may be further adapted to obtain a user feedback log, which includes all previous historical query results that take words in the first set of query keywords as search targets and all previous selection frequencies of the user on the historical query results.
[0055] The matching module may be further adapted to determine the second set of query keywords that matches the first set of query keywords according to the historical query results and selection frequencies in the user feedback log.
[0056] In implementation, the user-feedback-log obtaining module may be further adapted to obtain, as the selection frequencies, the selection frequencies on historical query results and/or the selection frequencies on the contents of historical query results.
[0057] FIG. 3 is a block diagram illustrating an embodiment of the matching module. As shown in the figure, the matching module includes: a content obtaining unit 2031, for obtaining the contents of historical query results corresponding to the first set of keywords; a word partitioning unit 2032, for carrying out word partitioning on the contents of historical query results, and obtaining words after word partitioning; and a matching unit 2033, for determining the second set of query keywords that matches the first set of query keywords according to the selection frequencies of the words after word partitioning.
[0058] In implementation, the word partitioning unit may also be further adapted to obtain, after word partitioning, words of one or more of the following types: words adjacent to the first set of query keywords after word partitioning; words containing the first set of query keywords after word partitioning; and words containing a component part of the first set of query keywords after word partitioning.
[0059] In implementation, the matching module may be further adapted to determine the second set of query keywords that matches the first set of query keywords according to the historical query results and selection frequencies, wherein the selection frequencies are larger than a given threshold.
[0060] FIG. 4 is a block diagram illustrating an embodiment of the first-query- keyword obtaining module. As shown in the figure, the first-query-keyword obtaining module may include: an information-content obtaining unit 201 1, for obtaining the information content that is input by a user; a word partitioning/decomposing unit 2012, for obtaining words after word partitioning after carrying out word partitioning on the information content, and/or decomposing the information content into characters; and a first- query-keyword determining unit 2013, for taking the words and/or characters after word partitioning as the first set of query keywords.
[0061] In some embodiments, the user-feedback-log obtaining module may be further adapted to obtain, as the selection frequencies, one or a combination of: click frequencies on the historical query results, display frequencies on the historical query results, reading time on the historical query results, and importance of the historical query results. [0062] In some embodiments, the first-query-keyword obtaining module may be further adapted to obtain the user characteristic of a user when the user inputs the first set of query keywords; the user-feedback-log obtaining module may be further adapted to obtain a user feedback log according to the user characteristic.
[0063] In some embodiments, the first-query-keyword obtaining module may be further adapted to obtain the user characteristic of a user when the user inputs the first set of query keywords;
[0064] the user-feedback-log obtaining module may also be further adapted to, when obtaining a user feedback log, obtain a user feedback log that includes historical query results that take the word to be matched as an object and the selection frequencies of the user on the historical query results, wherein the historical query results include the user characteristic.
[0065] In some embodiments, the first-query-keyword obtaining module may also be further adapted to obtain the user characteristic of a user when the user inputs the first set of query keywords. The matching module may be further adapted to determine the second set of query keywords according to the user characteristic when determining the second set of query keywords according to the user feedback log.
[0066] FIG.5 is a flowchart illustrating an embodiment of a process for word matching. As shown in the figure, it may include the following steps when word matching is performed: at 501, obtaining a word to be matched, such as the first set of query keywords; at 502, obtaining a user feedback log according to the word to be matched, wherein the user feedback log includes all previous historical query results that take the word to be matched as an object and all previous selection frequencies of users on the historical query results; at 503, determining words that match the word to be matched according to the historical query results and selection frequencies.
[0067] In some embodiments, the selection frequencies include: selection frequencies on historical query results and/or selection frequencies on the contents of historical query results.
[0068] In some embodiments, the process of determining words that match the word to be matched according to the selection frequencies on the contents of historical query results includes the following: obtaining the contents of historical query results of the word to be matched; carrying out word partitioning on the contents of historical query results, and obtaining words after word partitioning; and determining words that match the word to be matched according to the selection frequencies of the words after word partitioning.
[0069] In various embodiments, words after word partitioning refer to words of one or more of the following types: words adjacent to the word to be matched after word partitioning; words containing the word to be matched after word partitioning; and words containing a component part of the word to be matched after word partitioning. [0070] In some embodiments, when words that match the word to be matched are determined according to the historical query results and selection frequencies, the selection frequencies are larger than a given threshold.
[0071] The process of obtaining a keyword to be matched may be as follows: obtaining the information content that is input by a user; obtaining words after word partitioning after carrying out word partitioning on the information content and/or decomposing the information content into characters; and taking the words and/or characters after word partitioning as the word to be matched.
[0072] In some embodiments, the selection frequencies may include one or a combination of the following: click frequencies on the historical query results, display frequencies on the historical query results, reading time on the historical query results, and importance of the historical query results.
[0073] In some embodiments, the method may also further include the steps of obtaining the user characteristic of a user when the user inputs a word to be matched and obtaining a user feedback log according to the user characteristic when obtaining the user feedback log.
[0074] In some embodiments, the method may also further include the following steps: obtaining the user characteristic of a user when the user inputs a word to be matched; obtaining a user feedback log, wherein the user feedback log obtained includes historical query results that take the word to be matched as an object and the selection frequencies of users on the historical query results; and the historical query results include the user characteristic.
]0075] In some embodiments, the method may also further include the steps of obtaining the user characteristic of a user when the user inputs a word to be matched and determining words that match the word to be matched according to the user characteristic when determining words that match the word to be matched according to the user feedback log.
[0076] FIG.6 is a block diagram illustrating an embodiment of a device for word matching. As shown in the figure, the device includes: a word-to-be-matched obtaining module 601, for obtaining a word to be matched; a user- feedback-log obtaining module 602, for obtaining a user feedback log according to the word to be matched; and a matching module 603, for determining words that match the word to be matched according to the user feedback log. (0077] In some embodiments, the user-feedback-log obtaining module may be further adapted to obtain a user feedback log, which includes all previous historical query results that take the word to be matched as a target and all previous selection frequencies of users on the historical query results; and the matching module may be further adapted to determine words that match the word to be matched according to the historical query results and selection frequency in the user feedback log.
[0078] The user-feedback-log obtaining module may be further adapted to obtain, as the selection frequencies, the selection frequencies on historical query results and/or the selection frequencies on the contents of historical query results.
[0079] In some embodiments, the matching module includes: a content obtaining unit, for obtaining the contents of historical query results of the word to be matched; a word partitioning unit, for carrying out word partitioning on the contents of historical query results and obtaining words after word partitioning; and a matching unit, for determining words that match the word to be matched according to the selection frequencies of the words after word partitioning.
[0080] The word partitioning unit may be further adapted to obtain, after word partitioning, words in the following modes or a combination thereof: words adjacent to the word to be matched after word partitioning; words containing the word to be matched after word partitioning; and words containing a component part of the word to be matched after word partitioning.
[0081] The matching module may be further adapted to determine words that match the word to be matched according to the historical query results and selection frequencies, wherein the selection frequencies are larger than a given threshold.
[0082] The word-to-be-matched obtaining module may include: an information- content obtaining unit, for obtaining the information content that is input by a user; a word partitioning/decomposing unit, for obtaining words after word partitioning after carrying out word partitioning on the information content and/or decomposing the information content into characters; and a word-to-be-matched determining unit, for taking the words and/or characters after word partitioning as the word to be matched.
[0083] The user- feedback-log obtaining module may be further adapted to obtain, as the selection frequencies, one or a combination of: click frequencies on the historical query results, display frequencies on the historical query results, reading time on the historical query results, and importance of the historical query results.
[0084] In some embodiments, the word-to-be-matched obtaining module is further adapted to obtain the user characteristic of a user when the user inputs a word to be matched; the user- feedback-log obtaining module is further adapted to obtain a user feedback log according to the user characteristic.
[0085] In some embodiments, the word-to-be-matched obtaining module may be further adapted to obtain the user characteristic of a user when the user inputs a word to be matched and the user-feedback-log obtaining module may also be further adapted to, when obtaining a user feedback log, obtain a user feedback log that includes historical query results that take the word to be matched as an object and the selection frequencies of users on the historical query results, wherein the historical query results include the user characteristic.
[0086] In some embodiments, the word-to-be-matched obtaining module may also be further adapted to obtain the user characteristic of a user when the user inputs a word to be matched and the matching module may be further adapted to, when determining words that match the word to be matched according to the user feedback log, determine words that match the word to be matched according to the user characteristic.
[0087] It can be seen from the above embodiments that, based on the analysis of a user feedback log, the latent semantics of a word may be automatically found and the hidden relation between words can be found correctly. Further, the latent semantics of a word automatically found and the correlated semantics of a word to be queried may be used to improve the effect of the search engine. Moreover, when the latent word meaning of a query word is automatically found, a similar effect may also be obtained according to the word frequencies of separate words immediately adjacent to the query word in the search string, rather than only according to the word partitioning result. Therefore, in an embodiment of this application, the performance of the search engine may be improved via the latent word meaning of a word that is automatically found. In comparison with the traditional techniques, search accuracy and efficiency may be improved.
[0088] For easy description, the above system is divided into, according to functions, various modules or units, which are described respectively. However, during the implementation of the invention, the function of each module or unit may be realized in one or more software and/or hardware.
[0089] One skilled in the art should understand that, the embodiments of this application may be provided as a method, a system, or a computer program product. Therefore, this application may be in the form of full hardware embodiments, full software embodiments, or a combination thereof. Moreover, this application may be in the form of a computer program product that is implemented on one or more computer-usable storage media (including, without limitation, magnetic disk storage, CD-ROM and optical storage) containing computer-usable program codes.
[0090] This application is described referring to the flow chart and/or block diagram of the method, device (system) and computer program product according to the embodiments of this application. It should be understood that, each flow and/or block in the flow chart and/or block diagram and the combination of flow and/or block in the flow chart and/or block diagram may be realized via computer program instructions. Such computer program instructions may be provided to the processor of a general-purpose computer, special-purpose computer, a built-in processor or other programmable data processing devices, to produce a machine, so that the instructions executed by the processor of a computer or other programmable data processing devices may produce a device for realizing the functions specified in one or more flows in the flow chart and/or one or more blocks in the block diagram.
[0091] Such computer program instructions may also be stored in a computer- readable storage that can guide a computer or other programmable data processing devices to work in a specific mode, so that the instructions stored in the computer-readable storage may produce a manufacture including a commander equipment, wherein the commander equipment may realize the functions specified in one or more flows of the flow chart and one or more blocks in the block diagram.
[0092] Such computer program instructions may also be loaded to a computer or other programmable data processing devices, so that a series of operational processes may be executed on the computer or other programmable devices to produce a computer-realized processing, thereby the instructions executed on the computer or other programmable devices may provide a process for realizing the functions specified in one or more flows in the flow chart and/or one or more blocks in the block diagram.
[0093] Although preferred embodiments of this application have been described above, other variations and modifications can be made by one skilled in the art in the teaching of the basic creative conception. Therefore, the preferred embodiments and all these variations and modifications are intended to be contemplated by the appended claims.
[0094] It will be appreciated that one skilled in the art may make various modifications and alterations to the present invention without departing from the spirit and scope of the present invention. Accordingly, if these modifications and alterations to the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention intends to include all these modifications and alterations.
[0095] WHAT IS CLAIMED IS:

Claims

1. A method of performing a query, comprising: receiving a user input; determining a first set of query keywords based at least in part on the user input; obtaining, based on at least some of the first set of query keywords, a user feedback log that includes historical query results; determining a second set of query keywords based on the user feedback log; and making a query based on at least some of the second set of query keywords.
2. The method of Claim 1, further comprising: returning a query result; and updating a user feedback database with the query result.
3. The method of Claim 1 , wherein the user input is partitioned into a plurality of partitioned words based on context, and the first set of query keywords includes the plurality of partitioned words.
4. The method of Claim 3, wherein the first set of query keywords includes one or more words that are adjacent to at least some of the plurality of partitioned words, one or more words that include at least a portion of the plurality of partitioned words, one or more words that include at least a partitioned word in the plurality of partitioned words, or a combination thereof.
5. The method of Claim 1 , wherein the user feedback log includes a historical query result obtained using at least some of the first set of query keywords as a query target.
6. The method of Claim 1 , wherein the user feedback log includes a selection frequency.
7. The method of Claim 6, wherein the selection frequency is determined based at least in part on: click frequencies on the historical query results, display frequencies on the historical query results, reading time on the historical query results, importance of the historical query results, or a combination thereof.
8. The method of Claim 1 , wherein the second set of query keywords includes one or more latent words that correspond to the first set of query keywords, the latent words having latent meanings with respect to the first set of query keywords.
9. The method of Claim 1, wherein making the query based on at least some of the second set of query keywords includes sending the at least some of the second set of query keywords to a search engine.
10. The method of Claim 1, further comprising obtaining a user characteristic of a user 5 who generated the user input; and wherein the user feedback log is obtained at least in part according to the user characteristic.
11. The method of Claim 1 , further comprising obtaining a user characteristic of a user who generated the user input; and wherein the second set of query keywords is obtained at least in part according to the user characteristic. o
12. A query system comprising: an interface configured to receive a user input; one or more processors coupled to the interface, configured to: determine a first set of query keywords based at least in part on the user input; obtain, based on at least some of the first set of query keywords, a user5 feedback log that includes historical query results; determine second set of query keywords based on the user feedback log; and make a query based on at least some of the second set of query keywords; and one or more memories coupled to the one or more processors, configured to provide the processors with instructions. o
13. The system of Claim 12, further comprising: returning a query result; and updating a user feedback database with the query result.
14. The system of Claim 12, wherein the user input is partitioned into a plurality of partitioned words based on context, and the first set of query keywords includes the plurality5 of partitioned words.
15. The system of Claim 14, wherein the first set of query keywords includes one or more words that are adjacent to at least some of the plurality of partitioned words, one or more words that include at least a portion of the plurality of partitioned words, one or more words that include at least a partitioned word in the plurality of partitioned words, or a combination0 thereof.
16. The system of Claim 12, wherein the user feedback log includes a historical query result obtained using at least some of the first set of query keywords as a query target.
17. The system of Claim 12, wherein the user feedback log includes a selection frequency.
18. The system of Claim 17, wherein the selection frequency is determined based at least in part on: click frequencies on the historical query results, display frequencies on the historical query results, reading time on the historical query results, importance of the historical query results, or a combination thereof.
19. The system of Claim 12, wherein the second set of query keywords includes one or more latent words that correspond to the first set of query keywords, the latent words having latent meanings with respect to the first set of query keywords.
20. The system of Claim 12, wherein making the query based on at least some of the second set of query keywords includes sending the at least some of the second set of query keywords to a search engine.
21. The system of Claim 12, wherein the processor is further configured to obtain a user characteristic of a user who generated the user input; and wherein the user feedback log is obtained at least in part according to the user characteristic.
22. The system of Claim 12, wherein the processor is further configured to obtain a user characteristic of a user who generated the user input; and wherein the second set of query keywords is obtained at least in part according to the user characteristic.
PCT/US2010/000075 2009-01-12 2010-01-12 Search engine for refining context-based queries based upon historical user feedback WO2010080719A1 (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
JP2011545381A JP5679993B2 (en) 2009-01-12 2010-01-12 Method and query system for executing a query
EP10729396A EP2382534A4 (en) 2009-01-12 2010-01-12 Search engine for refining context-based queries based upon historical user feedback

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
CN200910001555.7 2009-01-12
CNA2009100015557A CN101464897A (en) 2009-01-12 2009-01-12 Word matching and information query method and device
US12/655,980 2010-01-11
US12/655,980 US8898180B2 (en) 2009-01-12 2010-01-11 Method and system for querying information

Publications (1)

Publication Number Publication Date
WO2010080719A1 true WO2010080719A1 (en) 2010-07-15

Family

ID=40805472

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2010/000075 WO2010080719A1 (en) 2009-01-12 2010-01-12 Search engine for refining context-based queries based upon historical user feedback

Country Status (5)

Country Link
US (2) US8898180B2 (en)
EP (1) EP2382534A4 (en)
JP (1) JP5679993B2 (en)
CN (1) CN101464897A (en)
WO (1) WO2010080719A1 (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103207881A (en) * 2012-01-17 2013-07-17 阿里巴巴集团控股有限公司 Query method and unit
US10275533B2 (en) 2015-06-18 2019-04-30 Baidu Online Network Technology (Beijing) Co., Ltd. Search engine and method for implementing the same
US11759514B2 (en) 2016-05-30 2023-09-19 Janssen Vaccines & Prevention B.V. Stabilized pre-fusion RSV F proteins

Families Citing this family (48)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100131513A1 (en) 2008-10-23 2010-05-27 Lundberg Steven W Patent mapping
US8666972B2 (en) * 2009-02-12 2014-03-04 International Business Machines Corporation System and method for content management and determination of search conditions
CN101833570A (en) * 2010-03-23 2010-09-15 深圳市五巨科技有限公司 Method and device for optimizing page push of mobile terminal
US8548981B1 (en) * 2010-06-23 2013-10-01 Google Inc. Providing relevance- and diversity-influenced advertisements including filtering
CN102567408B (en) 2010-12-31 2014-06-04 阿里巴巴集团控股有限公司 Method and device for recommending search keyword
US9904726B2 (en) 2011-05-04 2018-02-27 Black Hills IP Holdings, LLC. Apparatus and method for automated and assisted patent claim mapping and expense planning
US8880532B2 (en) 2011-06-29 2014-11-04 International Business Machines Corporation Interestingness of data
CN102855252B (en) * 2011-06-30 2015-09-09 北京百度网讯科技有限公司 A kind of need-based data retrieval method and device
US9940363B2 (en) 2011-10-03 2018-04-10 Black Hills Ip Holdings, Llc Systems, methods and user interfaces in a patent management system
CN103034665B (en) * 2011-10-10 2016-01-06 阿里巴巴集团控股有限公司 Information query method and device
US9043248B2 (en) 2012-03-29 2015-05-26 International Business Machines Corporation Learning rewrite rules for search database systems using query logs
US9313330B1 (en) * 2012-09-27 2016-04-12 West Corporation Identifying recorded call data segments of interest
US10395215B2 (en) * 2012-10-19 2019-08-27 International Business Machines Corporation Interpretation of statistical results
TW201435627A (en) * 2013-03-12 2014-09-16 Hon Hai Prec Ind Co Ltd System and method for optimizing search results
CN103207906B (en) * 2013-03-28 2016-12-28 百度在线网络技术(北京)有限公司 The offer method of Search Results and search engine
US20140317078A1 (en) * 2013-04-18 2014-10-23 SkillPages Holdings Limited Method and system for retrieving information
CN104346160B (en) * 2013-08-09 2018-02-27 联想(北京)有限公司 The method and electronic equipment of information processing
CN104636403B (en) * 2013-11-15 2019-03-26 腾讯科技(深圳)有限公司 Handle the method and device of inquiry request
CN104063454A (en) * 2014-06-24 2014-09-24 北京奇虎科技有限公司 Search push method and device for mining user demands
CN105389314B (en) * 2014-09-04 2019-07-23 中芯国际集成电路制造(上海)有限公司 A kind of journal file inquiry system and querying method
CN104281702B (en) * 2014-10-22 2017-07-11 国家电网公司 Data retrieval method and device based on electric power critical word participle
CN104462552B (en) * 2014-12-25 2018-07-17 北京奇虎科技有限公司 Question and answer page core word extracting method and device
CN104462553B (en) * 2014-12-25 2019-02-26 北京奇虎科技有限公司 Question and answer page relevant issues recommended method and device
CN104462556B (en) * 2014-12-25 2018-02-23 北京奇虎科技有限公司 Question and answer page relevant issues recommend method and apparatus
CN106844372B (en) * 2015-12-04 2021-09-24 菜鸟智能物流控股有限公司 Logistics information query method and device
CN107291685B (en) * 2016-04-13 2020-10-13 北京大学 Semantic recognition method and semantic recognition system
CN106227891A (en) * 2016-08-24 2016-12-14 广东华邦云计算股份有限公司 A kind of merchandise query short text semantic processes method based on pattern
CN108427686A (en) * 2017-02-15 2018-08-21 北京国双科技有限公司 Text data querying method and device
US10692157B2 (en) * 2017-03-28 2020-06-23 International Business Machines Corporation Selection of information sources based on social activities
CN107222526B (en) * 2017-05-16 2020-09-29 百度在线网络技术(北京)有限公司 Method, device and equipment for pushing promotion information and computer storage medium
CN107622054B (en) * 2017-09-26 2020-12-22 科大讯飞股份有限公司 Text data error correction method and device
CN110147426B (en) * 2017-12-01 2021-08-13 北京搜狗科技发展有限公司 Method for determining classification label of query text and related device
CN108984582B (en) * 2018-05-04 2023-07-28 中国信息安全研究院有限公司 Query request processing method
CN108897843B (en) * 2018-06-27 2021-12-24 吉安职业技术学院 Text self-adaptive recommendation method based on center method
CN109597986A (en) * 2018-10-16 2019-04-09 深圳壹账通智能科技有限公司 Localization method, device, equipment and the storage medium of abnormal problem
CN111324805B (en) * 2018-12-13 2024-02-13 北京搜狗科技发展有限公司 Query intention determining method and device, searching method and searching engine
CN109726226A (en) * 2019-01-03 2019-05-07 中国联合网络通信集团有限公司 Group's dialog box sort method and device
CN109933645B (en) * 2019-01-28 2024-01-23 平安科技(深圳)有限公司 Information query method, device, computer equipment and storage medium
CN109978498B (en) * 2019-03-15 2020-06-09 河北冀联人力资源服务集团有限公司 Task information processing method and device
US11281640B2 (en) * 2019-07-02 2022-03-22 Walmart Apollo, Llc Systems and methods for interleaving search results
CN111079421B (en) * 2019-11-25 2023-09-26 北京小米智能科技有限公司 Text information word segmentation processing method, device, terminal and storage medium
JP7127080B2 (en) * 2020-03-19 2022-08-29 ヤフー株式会社 Determination device, determination method and determination program
JP6948425B2 (en) * 2020-03-19 2021-10-13 ヤフー株式会社 Judgment device, judgment method and judgment program
CN112182193B (en) * 2020-10-19 2023-01-13 山东旗帜信息有限公司 Log obtaining method, device and medium in traffic industry
CN112925882A (en) * 2021-02-18 2021-06-08 联想(北京)有限公司 Information processing method and device
CN115077159A (en) * 2021-03-10 2022-09-20 松下电器研究开发(苏州)有限公司 Refrigerator and intelligent refrigerator system
US11860884B2 (en) * 2021-03-30 2024-01-02 Snap Inc. Search query modification database
CN113360743A (en) * 2021-07-09 2021-09-07 南方电网数字电网研究院有限公司 Query data filtering method and device, computer equipment and storage medium

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060129531A1 (en) * 2004-12-09 2006-06-15 International Business Machines Corporation Method and system for suggesting search engine keywords
US20080091670A1 (en) * 2006-10-11 2008-04-17 Collarity, Inc. Search phrase refinement by search term replacement
US20080222119A1 (en) * 2007-03-08 2008-09-11 Microsoft Corporation Detecting a user's location, local intent and travel intent from search queries

Family Cites Families (50)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH07219957A (en) 1994-01-28 1995-08-18 Matsushita Electric Ind Co Ltd Information sorting device, information retrieving device and information collecting device
US6567797B1 (en) 1999-01-26 2003-05-20 Xerox Corporation System and method for providing recommendations based on multi-modal user clusters
US6502091B1 (en) * 2000-02-23 2002-12-31 Hewlett-Packard Company Apparatus and method for discovering context groups and document categories by mining usage logs
JP2002092033A (en) 2000-09-12 2002-03-29 Sanyo Electric Co Ltd Information retrieving device
US6845374B1 (en) 2000-11-27 2005-01-18 Mailfrontier, Inc System and method for adaptive text recommendation
US7149804B2 (en) 2001-04-30 2006-12-12 Sony Computer Entertainment America Inc. Method and system for providing evaluation of text-based products
US7028024B1 (en) * 2001-07-20 2006-04-11 Vignette Corporation Information retrieval from a collection of information objects tagged with hierarchical keywords
US7092936B1 (en) 2001-08-22 2006-08-15 Oracle International Corporation System and method for search and recommendation based on usage mining
JP4003468B2 (en) * 2002-02-05 2007-11-07 株式会社日立製作所 Method and apparatus for retrieving similar data by relevance feedback
EP1860579A1 (en) * 2002-08-30 2007-11-28 Sony Deutschland Gmbh Method to split a multiuser profile
JP2005031949A (en) 2003-07-11 2005-02-03 Canon Inc Information retrieving method, information retrieving device, and program
US8086619B2 (en) * 2003-09-05 2011-12-27 Google Inc. System and method for providing search query refinements
US7451131B2 (en) 2003-12-08 2008-11-11 Iac Search & Media, Inc. Methods and systems for providing a response to a query
US7890526B1 (en) * 2003-12-30 2011-02-15 Microsoft Corporation Incremental query refinement
US7428529B2 (en) 2004-04-15 2008-09-23 Microsoft Corporation Term suggestion for multi-sense query
GB0414332D0 (en) * 2004-06-25 2004-07-28 British Telecomm Data storage and retrieval
US7707220B2 (en) * 2004-07-06 2010-04-27 Icosystem Corporation Methods and apparatus for interactive searching techniques
JP4461233B2 (en) 2005-03-30 2010-05-12 独立行政法人情報通信研究機構 MEDIA DATA SELECTION DEVICE, MEDIA DATA SELECTION METHOD, MEDIA DATA SELECTION PROGRAM, AND RECORDING MEDIUM CONTAINING THE PROGRAM
US8438142B2 (en) * 2005-05-04 2013-05-07 Google Inc. Suggesting and refining user input based on original user input
US8200687B2 (en) 2005-06-20 2012-06-12 Ebay Inc. System to generate related search queries
US7577665B2 (en) 2005-09-14 2009-08-18 Jumptap, Inc. User characteristic influenced search results
US7627548B2 (en) * 2005-11-22 2009-12-01 Google Inc. Inferring search category synonyms from user logs
US7689554B2 (en) * 2006-02-28 2010-03-30 Yahoo! Inc. System and method for identifying related queries for languages with multiple writing systems
CN100495398C (en) * 2006-03-30 2009-06-03 国际商业机器公司 Method for searching order in file system and correlation search engine
US8073860B2 (en) 2006-03-30 2011-12-06 Veveo, Inc. Method and system for incrementally selecting and providing relevant search engines in response to a user query
JP4958476B2 (en) 2006-05-24 2012-06-20 株式会社日立製作所 Search device
US7840538B2 (en) 2006-12-20 2010-11-23 Yahoo! Inc. Discovering query intent from search queries and concept networks
US7725453B1 (en) * 2006-12-29 2010-05-25 Google Inc. Custom search index
JP4909748B2 (en) 2007-01-22 2012-04-04 ヤフー株式会社 Method and server for recommending information related to Web page of reference source to Web page based on history information of page moving operation
JP2008293211A (en) 2007-05-23 2008-12-04 Hitachi Ltd Item recommendation system
US7734641B2 (en) 2007-05-25 2010-06-08 Peerset, Inc. Recommendation systems and methods using interest correlation
US7818320B2 (en) * 2007-05-31 2010-10-19 Yahoo! Inc. Enhanced search results based on user feedback relating to search result abstracts
CN101436186B (en) 2007-11-12 2012-09-05 北京搜狗科技发展有限公司 Method and system for providing related searches
JP4962967B2 (en) 2008-01-11 2012-06-27 ヤフー株式会社 Web page search server and query recommendation method
US8005828B2 (en) * 2008-02-05 2011-08-23 Yahoo! Inc. Learning query rewrite policies
CN101241512B (en) * 2008-03-10 2012-01-11 北京搜狗科技发展有限公司 Search method for redefining enquiry word and device therefor
US20090259646A1 (en) 2008-04-09 2009-10-15 Yahoo!, Inc. Method for Calculating Score for Search Query
US7890516B2 (en) * 2008-05-30 2011-02-15 Microsoft Corporation Recommending queries when searching against keywords
JP4869292B2 (en) 2008-06-20 2012-02-08 ヤフー株式会社 Server, method, and program for recommending search keywords
CN101751422A (en) 2008-12-08 2010-06-23 北京摩软科技有限公司 Method, mobile terminal and server for carrying out intelligent search at mobile terminal
CN101751437A (en) 2008-12-17 2010-06-23 中国科学院自动化研究所 Web active retrieval system based on reinforcement learning
JP2010146366A (en) 2008-12-19 2010-07-01 Toshiba Corp Information providing server
US8515966B2 (en) 2009-01-09 2013-08-20 Ebay Inc. Analyzing queries to generate product intention rules
US8719249B2 (en) 2009-05-12 2014-05-06 Microsoft Corporation Query classification
US8423568B2 (en) 2009-09-16 2013-04-16 Microsoft Corporation Query classification using implicit labels
CN102033877A (en) 2009-09-27 2011-04-27 阿里巴巴集团控股有限公司 Search method and device
US20110208730A1 (en) 2010-02-23 2011-08-25 Microsoft Corporation Context-aware searching
US9177045B2 (en) 2010-06-02 2015-11-03 Microsoft Technology Licensing, Llc Topical search engines and query context models
US8612432B2 (en) 2010-06-16 2013-12-17 Microsoft Corporation Determining query intent
US9465864B2 (en) 2010-09-29 2016-10-11 Excalibur Ip, Llc Training a search query intent classifier using wiki article titles and a search click log

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060129531A1 (en) * 2004-12-09 2006-06-15 International Business Machines Corporation Method and system for suggesting search engine keywords
US20080091670A1 (en) * 2006-10-11 2008-04-17 Collarity, Inc. Search phrase refinement by search term replacement
US20080222119A1 (en) * 2007-03-08 2008-09-11 Microsoft Corporation Detecting a user's location, local intent and travel intent from search queries

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103207881A (en) * 2012-01-17 2013-07-17 阿里巴巴集团控股有限公司 Query method and unit
CN103207881B (en) * 2012-01-17 2016-03-02 阿里巴巴集团控股有限公司 Querying method and device
US10275533B2 (en) 2015-06-18 2019-04-30 Baidu Online Network Technology (Beijing) Co., Ltd. Search engine and method for implementing the same
US11759514B2 (en) 2016-05-30 2023-09-19 Janssen Vaccines & Prevention B.V. Stabilized pre-fusion RSV F proteins

Also Published As

Publication number Publication date
US8898180B2 (en) 2014-11-25
US20150066888A1 (en) 2015-03-05
EP2382534A1 (en) 2011-11-02
EP2382534A4 (en) 2013-02-13
JP5679993B2 (en) 2015-03-04
US9430568B2 (en) 2016-08-30
US20100179948A1 (en) 2010-07-15
JP2012515379A (en) 2012-07-05
CN101464897A (en) 2009-06-24

Similar Documents

Publication Publication Date Title
US9430568B2 (en) Method and system for querying information
US8560513B2 (en) Searching for information based on generic attributes of the query
US7249121B1 (en) Identification of semantic units from within a search query
JP5575902B2 (en) Information retrieval based on query semantic patterns
US20100205198A1 (en) Search query disambiguation
WO2015124096A1 (en) Method and apparatus for determining morpheme importance analysis model
US20150278345A1 (en) Method, apparatus, and server for acquiring recommended topic
WO2016000555A1 (en) Methods and systems for recommending social network-based content and news
US8793120B1 (en) Behavior-driven multilingual stemming
US8825620B1 (en) Behavioral word segmentation for use in processing search queries
WO2021017306A1 (en) Personalized search method, system, and device employing user portrait, and storage medium
US8645363B2 (en) Spreading comments to other documents
EP2307951A1 (en) Method and apparatus for relating datasets by using semantic vectors and keyword analyses
CN107330057B (en) Optimization method and system for ElasticSearch search correlation algorithm
US20160306887A1 (en) Methods, apparatuses and systems for linked and personalized extended search
US10810245B2 (en) Hybrid method of building topic ontologies for publisher and marketer content and ad recommendations
CN104268175A (en) Data search device and method thereof
WO2016171709A1 (en) Text restructuring
US20200152081A1 (en) Food description processing methods and apparatuses
CN110245357B (en) Main entity identification method and device
CN104077320B (en) method and device for generating information to be issued
WO2016101737A1 (en) Search query method and apparatus
CN108509449B (en) Information processing method and server
CN110990701B (en) Book searching method, computing device and computer storage medium
TWI446191B (en) Word matching and information query method and device

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 10729396

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2011545381

Country of ref document: JP

Kind code of ref document: A

NENP Non-entry into the national phase

Ref country code: DE

WWE Wipo information: entry into national phase

Ref document number: 2010729396

Country of ref document: EP