CN102609458A - Method and device for picture recommendation - Google Patents
Method and device for picture recommendation Download PDFInfo
- Publication number
- CN102609458A CN102609458A CN2012100090437A CN201210009043A CN102609458A CN 102609458 A CN102609458 A CN 102609458A CN 2012100090437 A CN2012100090437 A CN 2012100090437A CN 201210009043 A CN201210009043 A CN 201210009043A CN 102609458 A CN102609458 A CN 102609458A
- Authority
- CN
- China
- Prior art keywords
- picture
- keyword
- speech
- text
- user
- Prior art date
- Legal status (The legal status 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 status listed.)
- Granted
Links
Images
Abstract
The invention provides a method and a device for picture recommendation. The method particularly includes: receiving a user's searching request, searching to obtain target pictures corresponding to the searching request and result pictures identical or similar to the target pictures; extracting key words describing semantic features of the picture according to the webpage text of the result pictures as the key words of the target pictures; matching the key words in a search log which records the target pictures corresponding to online search requests from users in the network and corresponding key words and recommending the target pictures matched with the key words to the user. The method and the device for picture recommendation are capable of providing pictures meeting personal demands of the users, and expanding access for information the users are interested in.
Description
Technical field
The application relates to the picture processing technology field, particularly relates to a kind of picture recommend method and device.
Background technology
At present along with the continuous development of network technology, the user no longer has been satisfied with just the search of text the requirement of search engine, and a lot of users also hope and can search for the network picture through search engine.
Present photographic search engine mostly adopts the text based search technique, and this technology is with the object of picture as database storing, and is described with key word.Yet for the visual signature that comprises in the picture, like color or shape etc., can't describe with text, like this, during the visual signature search pictures that comprises in need be according to picture, the text based search technique will be no longer suitable.For example, the user often runs into such problem, on website or computer, sees a picture that comprises article; But and do not know what the article in this picture are; So be difficult to the visual signature of these article is described in words out,, also be difficult in the picture that finds in the existing search engine with this picture analogies even if the good user of ability to express has described out with its visual signature; Cause search efficiency low, use network traffics bigger.
, use network traffics bigger problem low to above-mentioned search efficiency, some photographic search engines provide to scheme to search the figure function, should return to the user to scheme the searching figure function picture that vision content is consistent, to satisfy some search need of user.For example the certain user likes the collection picture, and least patient is exactly above the beautiful figure watermark to be arranged, as long as uploading pictures is clicked and just can be found the picture of not being with watermark to photographic search engine; And for example, can upload little picture, search out each version of this little picture, like clear big figure etc.
Also have some photographic search engines the picture recommendation function to be provided providing when scheming to search the figure function,, show the process flow diagram of picture recommend method in a kind of photographic search engine of prior art, specifically can comprise with reference to Fig. 1:
Visual signatures such as step 102, the color that extracts the inquiry picture, texture, shape;
The visual signature of picture carries out the similarity comparison in step 103, the visual signature that will inquire about picture and the database;
Because the comparison of visual signatures such as the color of picture recommendation results foundation, texture, shape obtains; So the similar main finger outward appearance of vision here is similar, but for example the user uploads the graceful picture in girl Jede, but the graceful hair color in girl Jede is golden in the picture; Then photographic search engine may return the similar picture that contains golden hair of vision; Like blondie's picture, sometimes even can return the picture of Cibotium barometz (L.) J. Sm, or the like.
But there are some individual demands in some user, like the picture that the user uploads Liu Dehua, also possibly hope to see pictures such as the film poster of Liu Dehua, individual description.At this moment, the consistent Search Results picture recommendation results similar with vision of vision content all can not satisfy user's individual demand in the prior art.
In a word, need the urgent technical matters that solves of those skilled in the art to be exactly: how the picture that agrees with users ' individualized requirement can be provided.
Summary of the invention
The application's technical matters to be solved provides a kind of picture recommend method and device, and the picture that agrees with users ' individualized requirement can be provided, the extending user information of interest obtain channel.
In order to address the above problem, the application discloses a kind of picture recommend method, comprising:
Receive user's query requests, and search obtains and corresponding Target Photo of this query requests and the as a result picture identical or approximate with this Target Photo;
Web page text according to said picture as a result place extracts the keyword of describing the picture semantic characteristic, as the keyword of this Target Photo;
In search log, carry out the coupling of keyword, and will recommend the user with the respective objects picture of keyword coupling; Said search log records the whole network user's online query requests corresponding Target Photo and corresponding keyword.
Preferably, the web page text at the said picture as a result of said foundation place, the step of the keyword of picture semantic characteristic is described in extraction, comprising:
According to the result that said web page text is carried out cluster analysis, remove web page text isolated in the said web page text, obtain remaining text;
It is the highest and have the speech or the phrase of practical significance to extract word frequency in the said residue text, as the keyword of describing the picture semantic characteristic.
Preferably, extract speech or the phrase that has practical significance in the said residue text through following steps:
Call the entity dictionary of structure in advance, when the entity speech in speech in said residue text or phrase and the said entity dictionary is complementary, keep institute's predicate or phrase; Said entity dictionary stores the entity speech with practical significance.
Preferably, extract speech or the phrase that has practical significance in the said residue text through following steps:
Extract speech or the phrase that has practical significance in the said residue text according to part of speech, said extraction process comprises:
When speech in said residue text or phrase are in interjection, pronoun or the tone auxiliary word any, abandon institute's predicate or phrase.
Preferably, the web page text at the said picture as a result of said foundation place, the step of the keyword of picture semantic characteristic is described in extraction, also comprises:
According to the adjacent co-occurrence frequency of other vocabulary in said keyword and the said residue text, add up edge speech adjacent in the said residue text with said keyword; With said edge speech with keyword as the keyword of describing the picture semantic characteristic.
Preferably, said method also comprises:
With corresponding each Target Photo of keyword coupling in the identical or approximate picture of filtering, obtain remaining picture;
Said will be that said residue picture is recommended the user with the step that the respective objects picture of keyword coupling is recommended the user.
Preferably, the said step that will recommend the user with the respective objects picture of keyword coupling comprises:
According to said search log, add up the corresponding on-line query request number of respective objects picture said and the keyword coupling;
Descending according to the on-line query request number will be recommended the user with the respective objects picture of keyword coupling.
Preferably, the picture that matees most for the query strategy corresponding of this Target Photo with this query requests; Said picture as a result is greater than other pictures of matching threshold except that Target Photo.
On the other hand, disclosed herein as well is a kind of picture recommendation apparatus, comprising:
The picture searching module is used to receive user's query requests, and search obtains and corresponding Target Photo of this query requests and the as a result picture identical or approximate with this Target Photo;
The keyword abstraction module is used for the web page text according to said picture as a result place, extracts the keyword of describing the picture semantic characteristic, as the keyword of this Target Photo;
Matching module is used for carrying out in search log the coupling of keyword; Said search log records the whole network user's online query requests corresponding Target Photo and corresponding keyword; And
The picture recommending module is used for recommending the user with the respective objects picture of keyword coupling.
Preferably, said keyword abstraction module comprises:
Remove submodule, be used for foundation, remove web page text isolated in the said web page text, obtain remaining text the result that said web page text carries out cluster analysis; And
Extract submodule, it is the highest and have the speech or the phrase of practical significance to be used for extracting said residue text word frequency, as the keyword of describing the picture semantic characteristic.
Preferably, said device also comprises:
The first practical significance abstraction module is used to call the entity dictionary of structure in advance, when the entity speech in speech in said residue text or phrase and the said entity dictionary is complementary, keeps speech or phrase in the said residue text; Said entity dictionary stores the entity speech with practical significance.
Preferably, said device also comprises:
The second practical significance abstraction module; Be used for extracting speech or the phrase that said residue text has practical significance according to part of speech; Said extraction process comprises: when speech in said residue text or phrase are in interjection, pronoun or the tone auxiliary word any, abandon speech or phrase in the said residue text.
Preferably, said keyword abstraction module also comprises:
Edge speech statistics submodule is used for the adjacent co-occurrence frequency according to said keyword and said other vocabulary of residue text, adds up edge speech adjacent with said keyword in the said residue text; With said edge speech with keyword as the keyword of describing the picture semantic characteristic.
Preferably, said device also comprises:
The filtering module, be used for the identical or approximate picture of corresponding each Target Photo filtering of keyword coupling, obtain remaining picture;
Said picture recommending module specifically is used for said residue picture is recommended the user.
Preferably, said picture recommending module comprises:
The number statistical submodule is used for according to said search log, adds up the corresponding on-line query request number of respective objects picture said and the keyword coupling;
Descending is recommended submodule, is used for according to the descending of on-line query request number corresponding respective objects picture with the keyword coupling being recommended the user.
Preferably, the picture that matees most for the query strategy corresponding of this Target Photo with this query requests; Said picture as a result is greater than other pictures of matching threshold except that Target Photo.
Compared with prior art, the application has the following advantages:
Describe the inquiry picture with respect to prior art employing visual signature, the application adopts keyword to describe the picture semantic characteristic of inquiry picture, and in search log, writes down the whole network on-line query request corresponding Target Photo and corresponding keyword; Because the described picture semantic characteristic of keyword can reflect user's hobby; Like this, when a submit queries request, the application can be according to the keyword of Target Photo in the keyword of resultant Target Photo and the said search log; Coupling obtains having the corresponding Target Photo of other user inquiring request of same interest hobby; Also promptly can agree with user's hobby with the respective objects picture of keyword coupling, therefore, the respective objects picture with the keyword coupling that will from search log, extract is recommended the active user; The picture that agrees with users ' individualized requirement is provided, has expanded the channel that obtains of user interest information.
Description of drawings
Fig. 1 is the process flow diagram of picture recommend method in a kind of photographic search engine of prior art;
Fig. 2 is the process flow diagram of a kind of picture recommend method of the application embodiment;
Fig. 3 is the structural drawing of a kind of picture recommendation apparatus of the application embodiment.
Embodiment
For above-mentioned purpose, the feature and advantage that make the application can be more obviously understandable, the application is done further detailed explanation below in conjunction with accompanying drawing and embodiment.
Users ' individualized requirement is derived from user's hobby often, and for example, certain user has the hobby of star-pursuing, and it is the bean vermicelli of Liu Dehua, then he when uploading the picture of Liu Dehua, pictures such as the very possible film poster of also hoping to see Liu Dehua, individual's description; And for example, another user is the moviegoer, and it has sincere hobby to " when happiness is knocked at the door " this film, and then he is when uploading the film poster of " when happiness is knocked at the door ", and other different placards of more these films are seen in very possible also hope.The similar Search Results of prior art vision is the users ' individualized requirement that can't satisfy under said circumstances.
One of core idea of the application embodiment is; The local feature of importing picture according to the active user obtains Target Photo and a plurality of as a result pictures similar or identical with its feature; The place of the picture as a result page is analyzed respectively; Word messages such as title, text in comprehensive each page, the keyword that obtains is related with Target Photo; Because the described picture semantic characteristic of keyword can reflect user's hobby; Like this; When a submit queries request, the application can be according to the keyword of Target Photo in the keyword of resultant Target Photo and the said search log, and coupling obtains having the corresponding corresponding Target Photo of other user inquiring request of same interest hobby; Also promptly can agree with user's hobby with the respective objects picture of keyword coupling; Therefore, will recommend the user with the respective objects picture of keyword coupling the picture that agrees with users ' individualized requirement can be provided, the extending user information of interest obtain channel.
With reference to Fig. 2, show the process flow diagram of a kind of picture recommend method of the application embodiment, specifically can comprise:
Step 201, receive user's query requests, and search obtains and corresponding Target Photo of this query requests and the as a result picture identical or approximate with this Target Photo;
The application can be applied in the photographic search engine, and in order to expand the function of photographic search engine, also promptly, it is original in to scheme to search the figure function to make that photographic search engine possesses, and possesses the application's picture recommendation function simultaneously.In fact, the application can also be applied to other search engine or searcher, and the application does not limit concrete applied environment.
In reality; The user can submit on-line query request in browser; The mode of the submission on-line query request here can comprise directly uploads local picture; The network address of picture perhaps is provided, and by the server automatic downloading picture, the application does not limit the mode of concrete submission on-line query request.Also promptly, among the application embodiment, can comprise the local picture that the user directly uploads, can comprise that also the network address of the picture that provides according to the user obtains picture with the direct corresponding picture of this query requests.
In concrete the realization; Server can be according to the direct vision content of corresponding picture of this query requests; Extract local feature, carry out picture searching then, mate with the local feature of each picture in the database; If matching rate in certain threshold range (as>90%), can think that the vision content of the two is consistent.
For the directly corresponding picture of this query requests and matching result, the two only has fine distinction, as whether with watermark, little picture and the difference etc. of picture greatly; Exclude these fine distinctions, the two is exactly identical picture.
Consider that the directly corresponding picture of this query requests possibly be poor qualities' such as the picture of band watermark or little picture picture; If with its storage object as search log; And the picture of finally recommending to the user is derived from search log; Like this, recommend to influence user's search experience with poor qualities' such as watermark or little picture figure sector-meeting to the user.Therefore, in a kind of preferred embodiment of the application, the picture that will the query strategy corresponding with query requests matees most is as Target Photo, and with the storage object of this Target Photo as search log.In reality, mate used database and often store some and be not with watermark and larger-size picture, like this, recommend can not improve user's search experience with watermark and larger-size picture to the user.
In a kind of preferred embodiment of the application; As a result picture be in the database except that Target Photo greater than other pictures of matching threshold, the conform to degree of degree that conform to of the query strategy that promptly picture is corresponding with query requests as a result less than the corresponding query strategy of Target Photo and query requests.In the present embodiment, the Target Photo that obtains and as a result picture sort by matching degree, the picture that matees most with query requests is a Target Photo, remaining picture as a result of picture by the matching degree displaying of sorting.In other embodiments; The corresponding result of user's query requests can sort by picture size or issuing time; Picture maximum or issue recently is as Target Photo with size, remaining picture as a result of picture by size by big to little or issuing time by near to the displaying of far sorting.Under normal conditions, picture and Target Photo only have fine distinction as a result, as whether with the difference of watermark, little picture and big picture etc.; Exclude these fine distinctions, the two is exactly identical picture.
Be appreciated that when the application is applied to photographic search engine server can also return to the user as Search Results with said picture as a result, to satisfy some search need of user.For example the certain user likes the collection picture, and least patient is exactly above the beautiful figure watermark to be arranged, as long as uploading pictures is clicked and just can be found the picture of not being with watermark to photographic search engine; And for example, can upload little picture, search out each version of this little picture, like clear big figure etc.
In a kind of applying examples of the application, said according to the direct vision content of corresponding picture of this query requests, the step that extracts local feature specifically can comprise:
At first, to this query requests directly the size of corresponding picture carry out normalization, oversize or too small picture is transformed within 640*640~300*300; The picture that uses two-dimentional local feature to detect after matrix and the normalization then carries out convolution operation; Moreover, the position that local extremum (maximal value and minimum value) point is wherein oriented in scanning in the picture after convolution; At last, according to the light and shade contrast of Local Extremum near zone, extract the directly local feature of corresponding picture of this query requests.Need to prove that in order to realize mating purpose, the size of picture after normalization that has identical original size with it in the directly corresponding picture of this query requests and the database should unanimity, for example, is all 300*300.
With reference to table 1, show the dimension of picture signal of a kind of normalization of the application front and back.
Table 1
In other embodiments, said picture as a result also can be searched in database for Target Photo is carried out feature extraction, matees resulting picture with the local feature of each picture in the database.
The web page text at step 202, the said picture as a result of foundation place extracts the keyword of describing the picture semantic characteristic, as the keyword of this Target Photo;
Because picture comes from network as a result, so can record each width of cloth web page text of picture as a result in the database of search engine or searcher, these web page texts generally include the text message of webpage, like page title, and the description text of picture periphery etc.
Because picture is the picture identical or approximate with Target Photo as a result, under normal conditions, the two only has fine distinction, as whether with the difference of watermark, little picture and big picture etc.; Exclude these fine distinctions, the two is exactly identical picture, that is to say, picture can be represented Target Photo fully as a result.
Like this, according to the web page text of picture as a result, the picture semantic characteristic that the keyword of extraction can the objective description Target Photo; And the picture semantic characteristic of Target Photo can reflect user's hobby to a certain extent, and for example, user search obtains the picture of Liu Dehua; Very possible this user of explanation is the bean vermicelli of Liu Dehua; And for example, user search obtains the film poster of " when happiness is knocked at the door ", explains that probably this user is the fan of " when happiness is knocked at the door " or the like.
In a kind of preferred embodiment of the application, the web page text at the said picture as a result of said foundation place, the step of the keyword of picture semantic characteristic is described in extraction, may further include:
Substep A1, said web page text is carried out cluster analysis;
Substep A2, according to cluster analysis result, remove in the said web page text isolated web page text, obtain remaining text;
In concrete the realization, can with every width of cloth as a result the web page text of picture be regarded as a document, to all as a result the web page text of picture carry out cluster analysis, those isolated texts that do not flock together are regarded as noise removal fall.The principle of cluster analysis is an arest neighbors binary tree cluster; When being applied to web page text; It is according to the repetition degree of web page text, be regarded as a class and merge repeating two parts of maximum web page texts, and the class after will merging is regarded as a web page text; The iteration repeat, until the repetition degree between two maximum web page texts of repeating can not reach merge threshold value till.
With reference to table 2 and table 3; Show the example of residue text after a kind of original web page text of the application and the cluster analysis respectively, wherein, the original web page text comprises the corresponding text of nine parts of webpages of 1-9; 2,4, the 9 noise texts of being numbered have wherein been removed in cluster analysis, obtain remaining text.
Table 2
Table 3
In the ideal case, the web page text of picture can be described the semantic content of corresponding picture truely and accurately as a result, and still, because the quality of web page text is uneven, at some in particular cases, the semantic content of web page text and picture is also uncorrelated.For example in the table 22,9 of original web page text.Though (text 4 is relevant with the picture semantic content, does not reach the merging threshold value with the repetition degree of other texts, therefore also has been removed.)
In reality; Picture as a result under the above-mentioned ideal situation is in the great majority, and picture as a result in particular cases is very indivedual, like this; The web page text of the picture as a result when cluster analysis under the ideal situation can flock together, and the web page text of picture as a result in particular cases is by isolated; Therefore, object or the incoherent isolated text of scene with in the Target Photo that above-mentioned cluster analysis can not flock together those are regarded as noise removal and fall, to improve the accuracy of keyword abstraction.
Substep A3, to extract word frequency in the said residue text the highest and have the speech or the phrase of practical significance, as the keyword of describing the picture semantic characteristic.
The application can provide the scheme that has the speech or the phrase of practical significance in the said residue text of following extraction:
Scheme one,
Can extract speech or the phrase that has practical significance in the said residue text through following steps:
Entity dictionary according to constructing in advance extracts the speech or the phrase that have practical significance in the said residue text, and said entity dictionary stores the entity speech with practical significance, and said extraction process can comprise:
Call the entity dictionary of structure in advance, when the entity speech in speech in said residue text or phrase and the said entity dictionary is complementary, keep speech or phrase in the said residue text.
The entity speech here mainly refers to represent the word of single or a plurality of entitative concepts, and it mainly comprises and is once called as noun, like name, movie name, item name etc.In reality, can collect the entity speech under the hobby classification, and construct corresponding entity dictionary in advance, the hobby classification here both can comprise; Amusement classifications such as film, TV, star, music, animation also can books, leisure classifications such as electronic product, clothes, shoes and hats etc.The application does not limit the make of concrete hobby classification and entity dictionary.
Scheme two,
Can extract speech or the phrase that has practical significance in the said residue text through following steps:
Extract speech or the phrase that has practical significance in the said residue text according to part of speech, said extraction process specifically can comprise:
When speech in said residue text or phrase are in interjection, pronoun or the tone auxiliary word any, abandon speech or phrase in the said residue text.
Because interjection, pronoun or tone auxiliary word etc. are everyday words, do not have practical significance usually, so when extracting, can carry out discard processing to it.Need to prove; Except interjection, pronoun or tone auxiliary word; This programme can also be according to actual conditions; Abandon the speech or the phrase of other part of speech in the said residue text, like adverbial word, preposition, conjunction, structure auxiliary word, dynamic any in auxiliary word, the onomatopoeia or the like, the application does not limit the part of speech that specifically abandons.
Need to prove; Extract the workload that has the speech or the phrase of practical significance in the said residue text in order to alleviate, in the application embodiment, preferably; Can at first extract speech or the phrase that word frequency is the highest in the said residue text and tentatively extracted the result; Then, from said preliminary extraction result, extract speech or phrase, finally extracted the result with practical significance.Certainly, those skilled in the art also can at first extract the speech or the phrase that have practical significance in the said residue text as required, extract the highest speech of word frequency or phrase then, and the application does not limit concrete succession.
In addition; Above-mentioned two kinds are extracted and to have the speech of practical significance in the said residue texts or the scheme of phrase can be used separately or be used in combination; Perhaps; Those skilled in the art can also adopt other to extract the scheme that has the speech or the phrase of practical significance in the said residue text according to actual needs, and the application does not limit this.
In the application's another kind of preferred embodiment, the web page text at the said picture as a result of said foundation place, the step of the keyword of picture semantic characteristic is described in extraction, can also comprise:
According to the adjacent co-occurrence frequency of other vocabulary in said keyword and the said residue text, add up edge speech adjacent in the said residue text with said keyword; With said edge speech with keyword as the keyword of describing the picture semantic characteristic.
Suppose that the user has uploaded the picture of Liu Dehua, and step 201-203 extracts keyword---" Liu Dehua " that obtains describing the picture semantic characteristic; In fact this user also hopes to see pictures such as the film poster of Liu Dehua, individual description; So; Can " Liu Dehua " be keyword, add up other more vocabulary of adjacent co-occurrence number of times with " Liu Dehua " in the said residue text, like " film ", " classical film ", " description " etc.; The keyword that like this, finally obtains can comprise: " Liu De China film ", " the classical film of Liu De China ", " Liu Dehua description " or the like.
Step 203, the coupling of keyword of in search log, carrying out, and will recommend the user with the respective objects picture of keyword coupling; Said search log records the whole network user's online query requests corresponding Target Photo and corresponding keyword.
Network operating system is designed with various journal files usually; Like application log; Security log, system journal or the like, when the user carried out certain operations on network system, these journal files can be noted some related contents of operation usually; The IP used (agreement that interconnects between the network, Internet Protocol), time, user name etc. like the user.
The application's search log generates to the whole network user's online query requests; Different with prior art is; Can in said search log, write down this on-line query request corresponding Target Photo and corresponding keyword; Wherein, said keyword obtains through execution in step 201-202.The whole network user here can comprise the user of internet; Also be that the user of internet is when submitting on-line query request in search engine or searcher; The server of search engine or searcher can generate corresponding search log; And the application can be from the internet server of all search engines or searcher collect search log, obtain search log.The application only stipulates the memory contents of search log, and can the obtain manner of concrete collection mode or search log not limited.
In concrete the realization; The search log of coupling institute foundation should be the whole network user's a search log; Obtain corresponding Target Photo to inquire other user inquirings that are complementary with the keyword of this Target Photo; The keyword that the keyword of the Target Photo that the coupling of the keyword here mainly writes down in the finger search log is identical with the keyword of current goal picture, comprise this Target Photo, or overlap each other, or the like.
The picture recommendation function that the application provides can satisfy users ' individualized requirement preferably; Because the described picture semantic characteristic of keyword can reflect user's hobby among the application; Like this; When a submit queries request; The application can be according to the keyword of Target Photo in the keyword of resultant Target Photo and the said search log, and coupling obtains having the corresponding Target Photo that other user inquiring of same interest hobby obtains, and also promptly can agree with user's hobby with the respective objects picture of keyword coupling.
In a kind of preferred embodiment of the application, before will recommending the user with the respective objects picture of keyword coupling, said method can also comprise:
With corresponding each Target Photo of keyword coupling in the identical or approximate picture of filtering, obtain remaining picture;
Said will with the respective objects picture of keyword coupling recommend the user step can for, said residue picture is recommended the user.
The front is mentioned, and in the ideal case, in each Target Photo corresponding with keyword, whether two identical or approximate width of cloth pictures only have identical or approximate with the nuances such as difference of watermark, little picture and big picture usually; In addition, corresponding each Target Photo with the keyword coupling is to obtain according to the keyword coupling of describing the picture semantic characteristic; Therefore, can think, if with corresponding each Target Photo of keyword coupling in have two width of cloth or identical or approximate picture more than two pairs, then not have the meaning of recommendation, so it is carried out filtering.
In the application's another kind of preferred embodiment, the said step that will recommend the user with the respective objects picture of keyword coupling may further include:
According to said search log, add up the corresponding on-line query request number of respective objects picture said and the keyword coupling;
Descending according to the on-line query request number is recommended the user with corresponding respective objects picture with the keyword coupling.
In some cases; Number said and corresponding each Target Photo that keyword matees possibly be big figure; More than 100 width of cloth; Whether the picture of these big figures agrees with users ' individualized requirement is difficult to expect, and needs to divide multipage that the picture of these big figures is shown in browser, makes the user need from multipage, extract own needed content.
This preferred embodiment is recommended the user according to the descending of on-line query request number with corresponding respective objects picture with the keyword coupling; The on-line query request number bright corresponding picture of the multilist user that had same interest hobby is more more paid close attention to; Also promptly, the application can preferentially recommend the high picture of the many attention rates of on-line query request number, therefore; The preferential picture of recommending can agree with users ' individualized requirement better, increases user's experience.
The application can provide the applying examples in the following scene:
Applying examples 1,
Step B1, receive the picture of the Liu Dehua that the user uploads, and obtain this picture corresponding Target Photo and the as a result picture identical or approximate with Target Photo with this picture search;
Step B2, according to the web page text at said picture as a result place, extract the keyword of describing the picture semantic characteristic, as the keyword of the picture of Liu Dehua, for example " Liu De China film ", " the classical film of Liu De China ", " Liu Dehua description " etc.;
Step B3, in search log, carry out the coupling of keyword; Obtain the corresponding Target Photo (like the picture concerned of more Liu De such as the film poster of Liu Dehua, individual description China) of reflection hobby that other users that like Liu De China equally upload, and recommend the user.
Applying examples 2,
" failing in love 33 days " film poster that step C1, reception user upload; Obtain corresponding Target Photo and the as a result picture identical or approximate through this film poster search with Target Photo;
Step C2, according to the web page text at said picture as a result place, extract the keyword of describing the picture semantic characteristic, as the keyword of this Target Photo, like " failing in love 33 days " etc.;
Step C3, in search log, carry out the coupling of keyword, obtain other and like the corresponding Target Photo (the for example different placards of this film) of reflection hobby that the user of " failing in love 33 days " this film uploads, and recommend the user.
For making those skilled in the art understand the application better, the process flow diagram of recommending the method example of star's picture in a kind of photographic search engine of the application below is provided, specifically can comprise:
But the graceful photo in girl Jede of the golden hair that step 1, reception user upload;
But step 2, from the graceful photo in girl Jede of golden hair, extract visual signature, compare, but obtain Target Photo consistent and picture as a result with the vision content of the graceful photo in girl Jede of golden hair with the visual signature of picture in the database;
Step 3, the web page text of picture is as a result carried out cluster analysis, those isolated texts that do not flock together are regarded as noise removal fall, it is the highest and have the speech or the phrase of practical significance to extract word frequency in the residue text, as keyword;
For example, table 4 shows a kind of word frequency example that remains in the text of the application.
Table 4
Speech | Word frequency in the text |
They | 12 |
But the girl Jede is graceful | 10 |
Film festival | 5 |
Venice | 4 |
?... | ... |
Wherein, " they " that word frequency is the highest do not have practical significance, and the keyword that therefore finally obtains is " but the girl Jede is graceful ".
Record corresponding Target Photo and the corresponding keyword of on-line query request that the whole network user submits in the search log of step 4, photographic search engine;
Step 5, be drawn into the keyword of Target Photo after, whether through keyword coupling (identical, comprise mutually or overlapping to some extent), inquiry obtains having with each keyword the Target Photo of semantic association;
Step 6, in filtering out Query Result with the identical or approximate Target Photo of the corresponding Target Photo of current inquiry after, remain the on-line query request number of picture in the statistics search log, and the part picture that the on-line query request number is maximum is recommended the user.
For example, when keyword was " but the girl Jede is graceful ", the on-line query request number of keyword that is associated with " but the girl Jede is graceful " in the search log and the corresponding picture of these keywords was as shown in table 4,
Table 4
Keyword | Corresponding Target Photo | Corresponding on-line query request number |
But beauty girl Jede is graceful | Picture a | 22 |
But the girl Jede is graceful | Picture b | 19 |
But the graceful stage photo in girl Jede | Picture c | 8 |
... | ?... | ... |
If for the user recommends 2 width of cloth pictures, the application will recommend picture a and picture b so.
At first, the said method example can extract keyword to describe the picture semantic characteristic of Target Photo, like star's in the photo name through photographic search engine.
Secondly, the said method example can provide the picture with identical semantic feature as content recommendation for the user, also promptly the picture materials that agrees with users ' individualized requirement can be provided.Like this, but when Target Photo is the girl Jede graceful photo of golden hair, can recommend different pictures, like the photo of these other hair colors of star about this star; Rather than only recommend the similar picture of vision.
Embodiment is corresponding with preceding method, and the application also provides a kind of picture recommendation apparatus, with reference to Fig. 3, specifically can comprise:
Picture searching module 301 is used to receive user's query requests, and search obtains and corresponding Target Photo of this query requests and the as a result picture identical or approximate with this Target Photo;
Keyword abstraction module 302 is used for the web page text according to said picture as a result place, extracts the keyword of describing the picture semantic characteristic, as the keyword of this Target Photo;
Matching module 303 is used for carrying out in search log the coupling of keyword; Said search log records the whole network user's online query requests corresponding Target Photo and corresponding keyword; And
Picture recommending module 304 is used for recommending the user with the respective objects picture of keyword coupling.
In the application embodiment, preferably, the picture that this Target Photo matees for the query strategy corresponding with this query requests most; In the database that said picture as a result is a server end except that Target Photo greater than other pictures of matching threshold.
In a kind of preferred embodiment of the application, said keyword abstraction module 302 may further include:
Remove submodule, be used for foundation, remove web page text isolated in the said web page text, obtain remaining text the result that said web page text carries out cluster analysis; And
Extract submodule, it is the highest and have the speech or the phrase of practical significance to be used for extracting said residue text word frequency, as the keyword of describing the picture semantic characteristic.
In the application's another kind of preferred embodiment, said device can also comprise:
The first practical significance abstraction module; Be used for the foundation entity dictionary of structure in advance; Extract the speech or the phrase that have practical significance in the said residue text; Said entity dictionary stores the entity speech with practical significance, and said extraction process comprises: when the entity speech in speech in said residue text or phrase and the said entity dictionary is complementary, keep speech or phrase in the said residue text.
In another preferred embodiment of the application, said device can also comprise:
The second practical significance abstraction module is used to call the entity dictionary of structure in advance, when the entity speech in speech in said residue text or phrase and the said entity dictionary is complementary, keeps speech or phrase in the said residue text; Said entity dictionary stores the entity speech with practical significance.
In a kind of preferred embodiment of the application, said keyword abstraction module 302 can also comprise:
Edge speech statistics submodule is used for the adjacent co-occurrence frequency according to said keyword and said other vocabulary of residue text, adds up edge speech adjacent with said keyword in the said residue text; With said edge speech with keyword as the keyword of describing the picture semantic characteristic.
In the application embodiment, preferably, said device can also comprise:
The filtering module, be used for the identical or approximate picture of corresponding each Target Photo filtering of keyword coupling, obtain remaining picture;
At this moment, said picture recommending module 304 can specifically be used for said residue picture is recommended the user.
In the application embodiment, preferably, said picture recommending module 304 specifically can comprise:
The number statistical submodule is used for according to said search log, adds up the corresponding on-line query request number of respective objects picture said and the keyword coupling; And
Descending is recommended submodule, is used for according to the descending of on-line query request number corresponding respective objects picture with the keyword coupling being recommended the user.
For device embodiment, because it is similar basically with method embodiment, so description is fairly simple, relevant part gets final product referring to the part explanation of method embodiment.
Each embodiment in this instructions all adopts the mode of going forward one by one to describe, and what each embodiment stressed all is and the difference of other embodiment that identical similar part is mutually referring to getting final product between each embodiment.
More than to a kind of picture recommend method and device that the application provided; Carried out detailed introduction; Used concrete example among this paper the application's principle and embodiment are set forth, the explanation of above embodiment just is used to help to understand the application's method and core concept thereof; Simultaneously, for one of ordinary skill in the art, according to the application's thought, the part that on embodiment and range of application, all can change, in sum, this description should not be construed as the restriction to the application.
Claims (16)
1. a picture recommend method is characterized in that, comprising:
Receive user's query requests, and search obtains and corresponding Target Photo of this query requests and the as a result picture identical or approximate with this Target Photo;
Web page text according to said picture as a result place extracts the keyword of describing the picture semantic characteristic, as the keyword of this Target Photo;
In search log, carry out the coupling of keyword, and will recommend the user with the respective objects picture of keyword coupling; Said search log records the whole network user's online query requests corresponding Target Photo and corresponding keyword.
2. the method for claim 1 is characterized in that, the web page text at the said picture as a result of said foundation place, and the step of the keyword of picture semantic characteristic is described in extraction, comprising:
According to the result that said web page text is carried out cluster analysis, remove web page text isolated in the said web page text, obtain remaining text;
It is the highest and have the speech or the phrase of practical significance to extract word frequency in the said residue text, as the keyword of describing the picture semantic characteristic.
3. method as claimed in claim 2 is characterized in that, extracts speech or the phrase that has practical significance in the said residue text through following steps:
Call the entity dictionary of structure in advance, when the entity speech in speech in said residue text or phrase and the said entity dictionary is complementary, keep institute's predicate or phrase; Said entity dictionary stores the entity speech with practical significance.
4. method as claimed in claim 2 is characterized in that, extracts speech or the phrase that has practical significance in the said residue text through following steps:
Extract speech or the phrase that has practical significance in the said residue text according to part of speech, said extraction process comprises:
When speech in said residue text or phrase are in interjection, pronoun or the tone auxiliary word any, abandon institute's predicate or phrase.
5. method as claimed in claim 2 is characterized in that, the web page text at the said picture as a result of said foundation place, and the step of the keyword of picture semantic characteristic is described in extraction, also comprises:
According to the adjacent co-occurrence frequency of other vocabulary in said keyword and the said residue text, add up edge speech adjacent in the said residue text with said keyword; With said edge speech with keyword as the keyword of describing the picture semantic characteristic.
6. like each described method in the claim 1 to 5, it is characterized in that said method also comprises:
With corresponding each Target Photo of keyword coupling in the identical or approximate picture of filtering, obtain remaining picture;
Said will be that said residue picture is recommended the user with the step that the respective objects picture of keyword coupling is recommended the user.
7. like each described method in the claim 1 to 5, it is characterized in that the said step that will recommend the user with the respective objects picture of keyword coupling comprises:
According to said search log, add up the corresponding on-line query request number of respective objects picture said and the keyword coupling;
Descending according to the on-line query request number will be recommended the user with the respective objects picture of keyword coupling.
8. like each described method in the claim 1 to 5, it is characterized in that the picture that this Target Photo matees for the query strategy corresponding with this query requests most; Said picture as a result is greater than other pictures of matching threshold except that Target Photo.
9. a picture recommendation apparatus is characterized in that, comprising:
The picture searching module is used to receive user's query requests, and search obtains and corresponding Target Photo of this query requests and the as a result picture identical or approximate with this Target Photo;
The keyword abstraction module is used for the web page text according to said picture as a result place, extracts the keyword of describing the picture semantic characteristic, as the keyword of this Target Photo;
Matching module is used for carrying out in search log the coupling of keyword; Said search log records the whole network user's online query requests corresponding Target Photo and corresponding keyword; And
The picture recommending module is used for recommending the user with the respective objects picture of keyword coupling.
10. device as claimed in claim 9 is characterized in that, said keyword abstraction module comprises:
Remove submodule, be used for foundation, remove web page text isolated in the said web page text, obtain remaining text the result that said web page text carries out cluster analysis; And
Extract submodule, it is the highest and have the speech or the phrase of practical significance to be used for extracting said residue text word frequency, as the keyword of describing the picture semantic characteristic.
11. device as claimed in claim 10 is characterized in that, also comprises:
The first practical significance abstraction module is used to call the entity dictionary of structure in advance, when the entity speech in speech in said residue text or phrase and the said entity dictionary is complementary, keeps speech or phrase in the said residue text; Said entity dictionary stores the entity speech with practical significance.
12. device as claimed in claim 10 is characterized in that, also comprises:
The second practical significance abstraction module; Be used for extracting speech or the phrase that said residue text has practical significance according to part of speech; Said extraction process comprises: when speech in said residue text or phrase are in interjection, pronoun or the tone auxiliary word any, abandon speech or phrase in the said residue text.
13. device as claimed in claim 10 is characterized in that, said keyword abstraction module also comprises:
Edge speech statistics submodule is used for the adjacent co-occurrence frequency according to said keyword and said other vocabulary of residue text, adds up edge speech adjacent with said keyword in the said residue text; With said edge speech with keyword as the keyword of describing the picture semantic characteristic.
14. like each described device in the claim 9 to 13, it is characterized in that, also comprise:
The filtering module, be used for the identical or approximate picture of corresponding each Target Photo filtering of keyword coupling, obtain remaining picture;
Said picture recommending module specifically is used for said residue picture is recommended the user.
15., it is characterized in that said picture recommending module comprises like each described device in the claim 9 to 13:
The number statistical submodule is used for according to said search log, adds up the corresponding on-line query request number of respective objects picture said and the keyword coupling;
Descending is recommended submodule, is used for according to the descending of on-line query request number corresponding respective objects picture with the keyword coupling being recommended the user.
16., it is characterized in that the picture that this Target Photo matees for the query strategy corresponding with this query requests most like each described device in the claim 9 to 13; Said picture as a result is greater than other pictures of matching threshold except that Target Photo.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201210009043.7A CN102609458B (en) | 2012-01-12 | 2012-01-12 | A kind of picture recommendation method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201210009043.7A CN102609458B (en) | 2012-01-12 | 2012-01-12 | A kind of picture recommendation method and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN102609458A true CN102609458A (en) | 2012-07-25 |
CN102609458B CN102609458B (en) | 2015-08-05 |
Family
ID=46526830
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201210009043.7A Active CN102609458B (en) | 2012-01-12 | 2012-01-12 | A kind of picture recommendation method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102609458B (en) |
Cited By (27)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102831176A (en) * | 2012-07-30 | 2012-12-19 | 东莞宇龙通信科技有限公司 | Method and server for recommending friends |
CN102902771A (en) * | 2012-09-27 | 2013-01-30 | 百度国际科技(深圳)有限公司 | Method, device and server for searching pictures |
CN103064903A (en) * | 2012-12-18 | 2013-04-24 | 厦门市美亚柏科信息股份有限公司 | Method and device for searching images |
CN103886506A (en) * | 2012-12-20 | 2014-06-25 | 联想(北京)有限公司 | Information processing method and electronic device |
CN103902678A (en) * | 2014-03-21 | 2014-07-02 | 百度在线网络技术(北京)有限公司 | Search recommendation method and device |
CN104090929A (en) * | 2014-06-23 | 2014-10-08 | 吕志雪 | Recommendation method and device of personalized picture |
CN104156694A (en) * | 2014-07-18 | 2014-11-19 | 百度在线网络技术(北京)有限公司 | Method and device for identifying target object of image |
CN104239457A (en) * | 2014-09-02 | 2014-12-24 | 百度在线网络技术(北京)有限公司 | Search recommending method and device |
CN104504111A (en) * | 2014-12-30 | 2015-04-08 | 百度在线网络技术(北京)有限公司 | Method and device for recommending image material |
CN104881451A (en) * | 2015-05-18 | 2015-09-02 | 百度在线网络技术(北京)有限公司 | Image searching method and image searching device |
CN104994404A (en) * | 2015-07-06 | 2015-10-21 | 无锡天脉聚源传媒科技有限公司 | Method and device for obtaining keywords for video |
CN105159938A (en) * | 2015-08-03 | 2015-12-16 | 百度在线网络技术(北京)有限公司 | Retrieval method and apparatus |
WO2016107190A1 (en) * | 2014-12-30 | 2016-07-07 | 百度在线网络技术(北京)有限公司 | Searching method and apparatus |
CN106791912A (en) * | 2016-12-28 | 2017-05-31 | Tcl集团股份有限公司 | A kind of poster picture match method and system |
WO2017181663A1 (en) * | 2016-04-21 | 2017-10-26 | 北京搜狗科技发展有限公司 | Method and device for matching image to search information |
CN107423318A (en) * | 2017-03-27 | 2017-12-01 | 北京珠穆朗玛移动通信有限公司 | A kind of method and mobile terminal of picture mark |
CN107862561A (en) * | 2017-09-15 | 2018-03-30 | 广州唯品会研究院有限公司 | A kind of method and apparatus that user-interest library is established based on picture attribute extraction |
CN107909088A (en) * | 2017-09-27 | 2018-04-13 | 百度在线网络技术(北京)有限公司 | Obtain method, apparatus, equipment and the computer-readable storage medium of training sample |
CN108027836A (en) * | 2015-09-24 | 2018-05-11 | 高通股份有限公司 | With the system and method for the query image access images of capture |
CN108230220A (en) * | 2016-12-15 | 2018-06-29 | 百度在线网络技术(北京)有限公司 | Watermark adding method and device |
CN110019675A (en) * | 2017-12-01 | 2019-07-16 | 北京搜狗科技发展有限公司 | A kind of method and device of keyword extraction |
WO2019150416A1 (en) * | 2018-01-30 | 2019-08-08 | 株式会社オプティム | Electronic data conversion system, method and program |
CN110134850A (en) * | 2018-02-09 | 2019-08-16 | 北京搜狗科技发展有限公司 | A kind of searching method and device |
CN110309324A (en) * | 2018-03-09 | 2019-10-08 | 北京搜狗科技发展有限公司 | A kind of searching method and relevant apparatus |
CN110941766A (en) * | 2019-12-10 | 2020-03-31 | 北京字节跳动网络技术有限公司 | Information pushing method and device, computer equipment and storage medium |
CN111831847A (en) * | 2020-07-08 | 2020-10-27 | 嘉兴智设信息科技有限公司 | Similar picture set recommendation method and system |
CN111859095A (en) * | 2019-04-02 | 2020-10-30 | 搜狗(杭州)智能科技有限公司 | Picture identification method and device |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20040267740A1 (en) * | 2000-10-30 | 2004-12-30 | Microsoft Corporation | Image retrieval systems and methods with semantic and feature based relevance feedback |
CN101241512A (en) * | 2008-03-10 | 2008-08-13 | 北京搜狗科技发展有限公司 | Search method for redefining enquiry word and device therefor |
CN102270234A (en) * | 2011-08-01 | 2011-12-07 | 北京航空航天大学 | Image search method and search engine |
-
2012
- 2012-01-12 CN CN201210009043.7A patent/CN102609458B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20040267740A1 (en) * | 2000-10-30 | 2004-12-30 | Microsoft Corporation | Image retrieval systems and methods with semantic and feature based relevance feedback |
CN101241512A (en) * | 2008-03-10 | 2008-08-13 | 北京搜狗科技发展有限公司 | Search method for redefining enquiry word and device therefor |
CN102270234A (en) * | 2011-08-01 | 2011-12-07 | 北京航空航天大学 | Image search method and search engine |
Cited By (41)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102831176B (en) * | 2012-07-30 | 2016-12-21 | 东莞宇龙通信科技有限公司 | The method of commending friends and server |
CN102831176A (en) * | 2012-07-30 | 2012-12-19 | 东莞宇龙通信科技有限公司 | Method and server for recommending friends |
CN102902771A (en) * | 2012-09-27 | 2013-01-30 | 百度国际科技(深圳)有限公司 | Method, device and server for searching pictures |
CN103064903A (en) * | 2012-12-18 | 2013-04-24 | 厦门市美亚柏科信息股份有限公司 | Method and device for searching images |
CN103064903B (en) * | 2012-12-18 | 2017-08-01 | 厦门市美亚柏科信息股份有限公司 | Picture retrieval method and device |
CN103886506A (en) * | 2012-12-20 | 2014-06-25 | 联想(北京)有限公司 | Information processing method and electronic device |
CN103886506B (en) * | 2012-12-20 | 2018-08-10 | 联想(北京)有限公司 | A kind of information processing method and electronic equipment |
CN103902678A (en) * | 2014-03-21 | 2014-07-02 | 百度在线网络技术(北京)有限公司 | Search recommendation method and device |
WO2015139487A1 (en) * | 2014-03-21 | 2015-09-24 | 百度在线网络技术(北京)有限公司 | Method and device for search and recommendation |
CN104090929A (en) * | 2014-06-23 | 2014-10-08 | 吕志雪 | Recommendation method and device of personalized picture |
CN104156694A (en) * | 2014-07-18 | 2014-11-19 | 百度在线网络技术(北京)有限公司 | Method and device for identifying target object of image |
CN104239457A (en) * | 2014-09-02 | 2014-12-24 | 百度在线网络技术(北京)有限公司 | Search recommending method and device |
CN104239457B (en) * | 2014-09-02 | 2019-03-12 | 百度在线网络技术(北京)有限公司 | Search for recommended method and device |
CN104504111A (en) * | 2014-12-30 | 2015-04-08 | 百度在线网络技术(北京)有限公司 | Method and device for recommending image material |
WO2016107190A1 (en) * | 2014-12-30 | 2016-07-07 | 百度在线网络技术(北京)有限公司 | Searching method and apparatus |
US10296541B2 (en) | 2014-12-30 | 2019-05-21 | Baidu Online Network Technology (Beijing) Co., Ltd. | Searching method and apparatus |
CN104504111B (en) * | 2014-12-30 | 2018-12-21 | 百度在线网络技术(北京)有限公司 | The recommended method and device of picture material |
CN104881451A (en) * | 2015-05-18 | 2015-09-02 | 百度在线网络技术(北京)有限公司 | Image searching method and image searching device |
CN104994404A (en) * | 2015-07-06 | 2015-10-21 | 无锡天脉聚源传媒科技有限公司 | Method and device for obtaining keywords for video |
CN105159938B (en) * | 2015-08-03 | 2018-11-30 | 百度在线网络技术(北京)有限公司 | Search method and device |
CN105159938A (en) * | 2015-08-03 | 2015-12-16 | 百度在线网络技术(北京)有限公司 | Retrieval method and apparatus |
CN108027836A (en) * | 2015-09-24 | 2018-05-11 | 高通股份有限公司 | With the system and method for the query image access images of capture |
WO2017181663A1 (en) * | 2016-04-21 | 2017-10-26 | 北京搜狗科技发展有限公司 | Method and device for matching image to search information |
CN108230220A (en) * | 2016-12-15 | 2018-06-29 | 百度在线网络技术(北京)有限公司 | Watermark adding method and device |
CN108230220B (en) * | 2016-12-15 | 2022-04-26 | 百度在线网络技术(北京)有限公司 | Watermark adding method and device |
CN106791912B (en) * | 2016-12-28 | 2020-07-03 | Tcl科技集团股份有限公司 | Poster picture matching method and system |
CN106791912A (en) * | 2016-12-28 | 2017-05-31 | Tcl集团股份有限公司 | A kind of poster picture match method and system |
CN107423318A (en) * | 2017-03-27 | 2017-12-01 | 北京珠穆朗玛移动通信有限公司 | A kind of method and mobile terminal of picture mark |
CN107862561A (en) * | 2017-09-15 | 2018-03-30 | 广州唯品会研究院有限公司 | A kind of method and apparatus that user-interest library is established based on picture attribute extraction |
CN107909088A (en) * | 2017-09-27 | 2018-04-13 | 百度在线网络技术(北京)有限公司 | Obtain method, apparatus, equipment and the computer-readable storage medium of training sample |
CN110019675B (en) * | 2017-12-01 | 2021-10-15 | 北京搜狗科技发展有限公司 | Keyword extraction method and device |
CN110019675A (en) * | 2017-12-01 | 2019-07-16 | 北京搜狗科技发展有限公司 | A kind of method and device of keyword extraction |
WO2019150416A1 (en) * | 2018-01-30 | 2019-08-08 | 株式会社オプティム | Electronic data conversion system, method and program |
CN110134850A (en) * | 2018-02-09 | 2019-08-16 | 北京搜狗科技发展有限公司 | A kind of searching method and device |
CN110309324A (en) * | 2018-03-09 | 2019-10-08 | 北京搜狗科技发展有限公司 | A kind of searching method and relevant apparatus |
CN110309324B (en) * | 2018-03-09 | 2024-03-22 | 北京搜狗科技发展有限公司 | Searching method and related device |
CN111859095A (en) * | 2019-04-02 | 2020-10-30 | 搜狗(杭州)智能科技有限公司 | Picture identification method and device |
CN110941766A (en) * | 2019-12-10 | 2020-03-31 | 北京字节跳动网络技术有限公司 | Information pushing method and device, computer equipment and storage medium |
CN110941766B (en) * | 2019-12-10 | 2023-10-20 | 北京字节跳动网络技术有限公司 | Information pushing method, device, computer equipment and storage medium |
CN111831847A (en) * | 2020-07-08 | 2020-10-27 | 嘉兴智设信息科技有限公司 | Similar picture set recommendation method and system |
CN111831847B (en) * | 2020-07-08 | 2023-08-18 | 嘉兴智设信息科技有限公司 | Similar picture set recommendation method and system |
Also Published As
Publication number | Publication date |
---|---|
CN102609458B (en) | 2015-08-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN102609458B (en) | A kind of picture recommendation method and device | |
US11238066B2 (en) | Generating personalized clusters of multimedia content elements based on user interests | |
US10860811B2 (en) | Method and device for generating review article of hot news, and terminal device | |
Begelman et al. | Automated tag clustering: Improving search and exploration in the tag space | |
CA2727537C (en) | System and method for compending blogs | |
US20090254540A1 (en) | Method and apparatus for automated tag generation for digital content | |
JP2017508214A (en) | Provide search recommendations | |
CN102007492A (en) | Method and apparatus for searching a plurality of stored digital images | |
Petkos et al. | Two-level Message Clustering for Topic Detection in Twitter. | |
JP2017535860A (en) | Method and apparatus for providing multimedia content | |
Zaharieva et al. | Automated social event detection in large photo collections | |
CA3217721A1 (en) | Commodity short title generation method and apparatus | |
CN104008180A (en) | Association method of structural data with picture, association device thereof | |
Nandi et al. | Bangla news recommendation using doc2vec | |
CN115905489B (en) | Method for providing bidding information search service | |
US20170185690A1 (en) | System and method for providing content recommendations based on personalized multimedia content element clusters | |
Liu et al. | Event analysis in social multimedia: a survey | |
CN106294358A (en) | The search method of a kind of information and system | |
Figueiredo et al. | Evidence of quality of textual features on the web 2.0 | |
CN103226601A (en) | Method and device for image search | |
Olivares et al. | Boosting image retrieval through aggregating search results based on visual annotations | |
Alam et al. | Labeling of web search result clusters using heuristic search and frequent Itemset | |
Choi et al. | Consento: a new framework for opinion based entity search and summarization | |
Vallet et al. | Exploiting external knowledge to improve video retrieval | |
JP2008204007A (en) | Image dictionary generation method, device and program |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant |