WO2008133368A1 - Information search ranking system and method based on users' attention levels - Google Patents
Information search ranking system and method based on users' attention levels Download PDFInfo
- Publication number
- WO2008133368A1 WO2008133368A1 PCT/KR2007/002701 KR2007002701W WO2008133368A1 WO 2008133368 A1 WO2008133368 A1 WO 2008133368A1 KR 2007002701 W KR2007002701 W KR 2007002701W WO 2008133368 A1 WO2008133368 A1 WO 2008133368A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- document
- attention
- action
- user
- rank
- Prior art date
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/951—Indexing; Web crawling techniques
Definitions
- the present invention relates to technology for calculating users' attention levels to a document on the basis of a User Action Log (UAL) and applying the users' attention levels to the ranking of search results .
- UAL User Action Log
- an information search system generates the results of a search by indexing documents matching a keyword entered by a user.
- Documents included in the results of a search are provided in the form of a ranked list (a set of links indicating documents) through a statistical technique, such as content analysis or link analysis .
- a document such as a web page, includes content and metadata.
- Content has an inclusive meaning including audio and video files as well as text.
- Metadata may include various attributes, such as a document language, a document title, a document size, a document identifier (for example, Uniform Resource Locator [URL] information) , a document format, category, and other attributes.
- the content and metadata of documents, and information about the relation information between documents are generally used.
- information is described from the standpoint of an information provider who creates content or desires to distribute content, so that a description made from the standpoint of end consumers who consume content is not considered.
- user-centered information such as attractive content or currently popular content, is excluded from the factors determining ranking, but only provider-centered factors, such as document titles or backlinks, are used.
- a representative information search system is 'Google' .
- a search process is performed in such a way that link information (hyperlinks) , indicating a given document, is analyzed in addition to information included in the document on the basis of a 'PageRank' technique, a PageRank value is assigned to the given document, and analysis information contained in the document and the above assigned PageRank value (ranging from 0 to 10) are summed, and thus ranked search results are provided.
- link information hyperlinks
- PageRank value ranging from 0 to 10
- an object of the present invention is to provide an information search ranking system and method, which convert actions taken by a user on each individual document (content) into an attention level, assign the attention level to a given document, and apply the assigned attention level to the ranking of search results, thus providing improved search results to the user.
- the present invention provides an information search ranking method fundamentally applied to a system including a search engine for searching documents, stored in a document DB, for a desired document in response to a search request and providing ranked search results .
- the information search ranking method based on users' attention levels comprises a step of collecting a plurality of documents over an information network and storing the documents in a document DB, a step of collecting and storing User Action Logs (UALs) from at least one user terminal or at least one network provider server, a step of calculating an Attention Rank (AR) from a multiplication of an Attention Value (AV) for all actions taken by each user by an Influence Value (IV) of the user for each action, with respect to all users who access an individual stored document, on a basis of the collected UALs, and storing the calculated AR in an attention rank DB, and a step of calculating a Rank Value (RV) with reference to the AR stored in the attention rank DB, with respect to each document searched in the document DB in response to a keyword-based search request received from the user terminal, thus providing ranked search results .
- UALs User Action Logs
- AR Attention Rank
- AR Attention Rank
- AV Attention Value
- IV Influence Value
- the Attention Value (AV) may be calculated by multiplying a predetermined weight ( w k ) , assigned to each action taken by a certain user accessing the individual document, by a sigmoid function that uses an elapsed time from the action (t) as a variable.
- the Influence Value (IV) may be calculated by the following equation:
- ⁇ is a sigmoid function
- c h is a total number of all actions taken by a user h
- m is a value obtained by dividing a sum of values c h by a total number of users .
- the present invention provides an information search ranking system and method, which use actions of users taken on an individual document as users' attention levels based on the memory model of a human being, thus providing excellent ranked search results sensitive to the users' attention or preference in response to a keyword- based search request. Further, the present invention can provide excellent ranked search results, even for the recent rapid proliferation of User Created Content (UCC) .
- UCC User Created Content
- FIG. 1 is a diagram showing the overall system to which the technical spirit of the present invention is applied;
- FIG. 2 is a diagram showing the construction of an information search ranking system according to the present invention.
- FIG. 3 is a diagram showing the detailed construction of an attention rank calculation module according to the present invention.
- FIG. 4 is a flowchart of an information search ranking method according to the present invention.
- FIG. 1 is a diagram showing the overall system to which the present invention is applied.
- the system includes a document-using user terminal 100 for viewing documents provided by one or more content (document) servers (not shown) connected to an information network, and creating a User Action Log (UAL) for the viewed document, an information search ranking system 200 for collecting and indexing a plurality of documents over the information network, collecting UALs from the document-using user terminal or a network provider server (not shown) , calculating an Attention Value (AV) on the basis of the collected UALs, and assigning an Attention Rank (AR) to each collected document, and a document-search user terminal 300 for requesting the information search ranking system 200 to search for a keyword, and receiving ranked search results, in which the user's attention level is taken into account, from the information search ranking system 200.
- a document-using user terminal 100 for viewing documents provided by one or more content (document) servers (not shown) connected to an information network, and creating a User Action Log (UAL) for the viewed document
- an information search ranking system 200 for collecting and indexing a plurality of documents over the information network, collecting
- user terminals include mobile phones or computers enabling Internet communication, and are classified into a 'document-using user terminal' and a 'document-search user terminal' for convenience of description.
- the 'document-using user terminal' is defined based on the function of viewing documents (for example, web surfing) provided in a plurality of content servers connected to the information network, and creating a UAL based on the viewed documents.
- the 'document-search user terminal' is defined based on the function of storing Attention Ranks (ARs) for respective documents on the basis of a plurality of UALs collected by the information search ranking system 200, and then receiving ranked search results matching a keyword from the information search ranking system 200.
- ARs Attention Ranks
- both the user terminals can be considered to be the same 'user terminal' .
- 'User Action Log means a log file obtained by recording actions taken by the user (user actions) when the document-using user terminal views a certain document, and includes 1) a document identifier, 2) an action identifier, 3) an action type, 4) an action time, and 5) supplementary data.
- the 'document identifier' is an identifier for a target document on which an action is taken.
- the 'action identifier' is information for identifying a user who took an action, and may be, for example, an IP address or a Medium Access Control (MAC) address when the user uses a computer, or may be, for example, the phone number or unique information of a mobile phone when the user uses a mobile phone .
- MAC Medium Access Control
- the 'action type' includes various types of actions, such as view (action of viewing summary information about a document, previewing a document, etc.), play (action of playing video, music, images, etc.), detailed view (action of viewing the full text of a document rather than summary information) , save (action of saving or storing a document) , buy (action of purchasing a document or content for sale) , recommend (action of recommending a document or content to another person) , evaluate (action of representing an individual's opinion about a document or content as digitized or standardized information) , attach supplementary information (action of attaching supplementary information, such as a comment or tag, to a document or content) , and bookmark
- the 'supplementary data' may include environmental information about the action taken by the user, for example, the location of the user, the environmental conditions (travel) , etc.
- FIG. 2 is a diagram showing the detailed construction of the information search ranking system 200 according to an embodiment of the present invention.
- the information search ranking system 200 includes a document collection module 210, a search engine 220, a UAL collection module 230, an attention value calculation module 240, a rank value calculation module 250, a document DB 10, a UAL DB 20, and an attention rank DB 30.
- the document collection module 210 collects a plurality of documents over the information network, and stores the documents in the document DB 10 so that the documents are indexed and stored in order to respond to a keyword-based search request (a search request received from the document- search user terminal) .
- the search engine 220 is an engine provided with functions required for a typical search operation of searching the document DB 10 for an input keyword and providing search results.
- the search engine 220 provides ranked search results while referring to the Attention Ranks (ARs) of the attention rank DB 30, respectively related to the documents constituting the search results, assigns Rank Values (RVs) , in which the 'users' attention levels' are taken into account, to the documents, and provides the documents, to which the RVs are assigned, in the form of a ranked list. This procedure is described in detail below.
- the 'user's attention level' is a principal factor for evaluating documents (content) , and is obtained by digitizing a user' s attention level using a memory model of the user.
- a human being recognizes a certain fact or an object, and then gradually forgets the fact or the object as time elapses.
- This is the memory model of a human being, represented by a sigmoid function.
- link information the PageRank of Google
- the total number of hits for a given document is generally used as an index.
- such an index is only a simple accumulative index, which does not take into account the temporal component of the memory model .
- the UAL collection module 230 collects User Action Logs (UALs) from the document-using user terminal 100 or the network provider server, and stores and manages the UALs in the UAL DB 20.
- the UALs can be collected from a logging tool (software organized in a web browser) installed in the document-using user terminal 100, or from UALs stored in the network provider server. It will be apparent that the two collection methods can be used together.
- the attention rank calculation module 240 includes the attention value calculation unit 242 and the action influence value calculation unit 244, thus calculating an Attention Rank (AR) from the multiplication of an Attention Value (AV) for all actions taken by each user by the Influence Value (IV) of the user for each action, with respect to all users who access an individual document (p of the following Equation [1] ) , and updating the calculated AR in the attention rank DB 30.
- the Attention Rank (AR) is defined by the following Equation [1] .
- AR(p) ⁇ [AV(e hpk )xIV(h)] [1] forall h,k
- Equation [1] the AV, calculated by the attention value calculation unit 242, is represented by the following Equation [2] , where e h pk means that a user h takes an action I on a document p , w k is a predetermined weight previously assigned to the type of action k , t is the difference between the time at which the AV is calculated and the time at which the action k actually occurs, that is, the elapsed time from the action k (hereinafter referred to as an 'elapsed time from an action' ) .
- the weight has a value ranging from 0 to 1 according to the type of action.
- a weight can be defined as 0.2, for 'play', as 0.4, for 'detailed view', as 0.5, and for 'recommend', as 0.9.
- Equation [2] is a sigmoid function having the elapsed time from an action t as a variable, as defined in the right term thereof, and is implemented by modeling the memory of a human being, as described above. Therefore, the weight w k decreases as the elapsed time from the action increases.
- the influence value (IV) of the user who took the action must be taken into account in the AV calculated by the attention value calculation unit 242. For example, if it is assumed that there is a user who took 1,000 actions and a user who took 10 actions, it is determined that the former user participates in the document more than the latter user by a factor of 100 times. In other words, when users' attention levels paid to the corresponding document are calculated, the user who took 1,000 actions has an excessively large influence on the document, compared to the remaining users. Therefore, a correction that allows the influence value (IV) of the user calculated by the action influence value calculation unit 244 to be taken into account in the AV must be performed.
- the influence value (IV) of the user for a given action is represented by the following Equation [3] ,
- ⁇ is the sigmoid function
- c h is the total number of all actions taken by the user h
- m is the value obtained by dividing the sum of values c h by the total number of users
- Equation [3] can be summarized as number of users
- the rank value calculation module 250 calculates a Rank Value (RV) with reference to the AR for each extracted document, and thus ranks the documents.
- RV Rank Value
- the AR can be used as a factor for calculating the RV, but the degree of relation of each document to a keyword can be additionally calculated using the following Equation [4] according to the circumstances,
- RV(p) DocRel(p)xAR(p) [4]
- DocRel is the degree of relation of a predetermined document to the keyword.
- the document collection module 210 collects and stores a plurality of documents over the information network at step SlOO . This step is performed by a typical information search system at predetermined periods .
- the UAL collection module 230 collects UALs for documents from the document-using user terminal 100 or the network provider server, and stores the UALs at step S200.
- the attention rank calculation module 240 calculates an AR in consideration of both the AV for all actions, taken by each user, and the IV of the user for each action, with respect to all users who took actions on each document (each document stored in the document DB) , on the basis of the UALs collected at step S200, at step S300.
- the search engine 220 searches the document DB 10 for documents matching a keyword at step S400.
- the rank value calculation module 250 calculates the Rank Values (RVs) of found documents with reference to ARs for respective found documents stored in the attention rank DB 30, and ranks the found documents at step S500.
- RVs Rank Values
- the search engine 220 provides the ranked search results to the document-search user terminal 300 that requested the search at step S600. Through a series of these steps, a querying person (user) can view excellent search results, in which the 'user's attention levels' are taken into account .
- the present invention provides an information search ranking system and method, which use actions of users taken on an individual document as users' attention levels based on the memory model of a human being, thus providing excellent ranked search results sensitive to the users' attention or preference in response to a keyword- based search request. Further, the present invention can provide excellent ranked search results, even for the recent rapid proliferation of User Created Content (UCC) .
- UCC User Created Content
Abstract
The present invention relates to an information search ranking system and method based on users' attention levels. The information search ranking system includes a UAL collection module (230) collects a plurality of UALs from at least one user terminal or at least one network provider server. An attention rank calculation module (240) calculates an Attention Rank (AR) from a multiplication of an Attention Value (AV) for all actions taken by each user by an Influence Value (IV) of the user for each action, with respect to all users who access an individual document, on a basis of the collected UALs, and updates the calculated AR. A rank value calculation module (250) calculates a Rank Value (RV) with reference to the AR, with respect to each document searched by the search engine in the document DB in response to a keyword-based search request, thus ranking the documents.
Description
[DESCRIPTION]
[invention Title]
INFORMATION SEARCH RANKING SYSTEM AND METHOD BASED ON USERS' ATTENTION LEVELS
[Technical Field]
The present invention relates to technology for calculating users' attention levels to a document on the basis of a User Action Log (UAL) and applying the users' attention levels to the ranking of search results .
[Background Art]
As well known to those skilled in the art, an information search system generates the results of a search by indexing documents matching a keyword entered by a user. Documents included in the results of a search are provided in the form of a ranked list (a set of links indicating documents) through a statistical technique, such as content analysis or link analysis .
In this case, the term 'document' is popularly used with a somewhat vague meaning, but can be typically defined as a web page. A document, such as a web page, includes content and metadata. Content has an inclusive meaning including audio and video files as well as text. Metadata may include various attributes, such as a document language, a document title, a document size, a document identifier (for example,
Uniform Resource Locator [URL] information) , a document format, category, and other attributes.
Meanwhile, in the ranking of information searches, the content and metadata of documents, and information about the relation information between documents (for example, link or category) are generally used. However, such information is described from the standpoint of an information provider who creates content or desires to distribute content, so that a description made from the standpoint of end consumers who consume content is not considered. For example, user-centered information, such as attractive content or currently popular content, is excluded from the factors determining ranking, but only provider-centered factors, such as document titles or backlinks, are used. A representative information search system is 'Google' . In the case of 'Google' , a search process is performed in such a way that link information (hyperlinks) , indicating a given document, is analyzed in addition to information included in the document on the basis of a 'PageRank' technique, a PageRank value is assigned to the given document, and analysis information contained in the document and the above assigned PageRank value (ranging from 0 to 10) are summed, and thus ranked search results are provided. Such a technique can be effectively realized in web sites in which link information is the main factor.
However, User Created Content (UCC) and mobile content, for example, video content, blog content, etc., which is
proliferating rapidly these days, have insufficient link information, and thus satisfactory search results cannot be predicted using only the PageRank technique.
[Disclosure] [Technical Problem]
Accordingly, the present invention has been made keeping in mind the above problems occurring in the prior art, and an object of the present invention is to provide an information search ranking system and method, which convert actions taken by a user on each individual document (content) into an attention level, assign the attention level to a given document, and apply the assigned attention level to the ranking of search results, thus providing improved search results to the user.
[Technical Solution]
In order to accomplish the above object, the present invention provides an information search ranking method fundamentally applied to a system including a search engine for searching documents, stored in a document DB, for a desired document in response to a search request and providing ranked search results .
The information search ranking method based on users' attention levels according to the present invention comprises a step of collecting a plurality of documents over an information network and storing the documents in a document
DB, a step of collecting and storing User Action Logs (UALs) from at least one user terminal or at least one network provider server, a step of calculating an Attention Rank (AR) from a multiplication of an Attention Value (AV) for all actions taken by each user by an Influence Value (IV) of the user for each action, with respect to all users who access an individual stored document, on a basis of the collected UALs, and storing the calculated AR in an attention rank DB, and a step of calculating a Rank Value (RV) with reference to the AR stored in the attention rank DB, with respect to each document searched in the document DB in response to a keyword-based search request received from the user terminal, thus providing ranked search results .
Preferably, the Attention Value (AV) may be calculated by multiplying a predetermined weight ( wk ) , assigned to each action taken by a certain user accessing the individual document, by a sigmoid function that uses an elapsed time from the action (t) as a variable.
Preferably, the Influence Value (IV) may be calculated by the following equation:
where σ is a sigmoid function, ch is a total number of all actions taken by a user h , and m is a value obtained by dividing a sum of values ch by a total number of users .
[Advantageous Effects]
Accordingly, the present invention provides an information search ranking system and method, which use actions of users taken on an individual document as users' attention levels based on the memory model of a human being, thus providing excellent ranked search results sensitive to the users' attention or preference in response to a keyword- based search request. Further, the present invention can provide excellent ranked search results, even for the recent rapid proliferation of User Created Content (UCC) .
[Description of Drawings]
FIG. 1 is a diagram showing the overall system to which the technical spirit of the present invention is applied;
FIG. 2 is a diagram showing the construction of an information search ranking system according to the present invention;
FIG. 3 is a diagram showing the detailed construction of an attention rank calculation module according to the present invention; and
FIG. 4 is a flowchart of an information search ranking method according to the present invention.
[Best Mode]
The above and other objects, features and other advantages of the present invention will be more clearly understood from the following detailed description taken in conjunction with the accompanying drawings. It should be noted that detailed
descriptions may be omitted if it is determined that the detailed descriptions of related well-known functions and construction may make the gist of the present invention unclear. FIG. 1 is a diagram showing the overall system to which the present invention is applied. The system includes a document-using user terminal 100 for viewing documents provided by one or more content (document) servers (not shown) connected to an information network, and creating a User Action Log (UAL) for the viewed document, an information search ranking system 200 for collecting and indexing a plurality of documents over the information network, collecting UALs from the document-using user terminal or a network provider server (not shown) , calculating an Attention Value (AV) on the basis of the collected UALs, and assigning an Attention Rank (AR) to each collected document, and a document-search user terminal 300 for requesting the information search ranking system 200 to search for a keyword, and receiving ranked search results, in which the user's attention level is taken into account, from the information search ranking system 200.
In the present invention, user terminals include mobile phones or computers enabling Internet communication, and are classified into a 'document-using user terminal' and a 'document-search user terminal' for convenience of description. The 'document-using user terminal' is defined based on the function of viewing documents (for example, web
surfing) provided in a plurality of content servers connected to the information network, and creating a UAL based on the viewed documents. The 'document-search user terminal' is defined based on the function of storing Attention Ranks (ARs) for respective documents on the basis of a plurality of UALs collected by the information search ranking system 200, and then receiving ranked search results matching a keyword from the information search ranking system 200. However, both the user terminals can be considered to be the same 'user terminal' .
Meanwhile, the term 'User Action Log (UAL) ' means a log file obtained by recording actions taken by the user (user actions) when the document-using user terminal views a certain document, and includes 1) a document identifier, 2) an action identifier, 3) an action type, 4) an action time, and 5) supplementary data.
In detail, the 'document identifier' is an identifier for a target document on which an action is taken. The 'action identifier' is information for identifying a user who took an action, and may be, for example, an IP address or a Medium Access Control (MAC) address when the user uses a computer, or may be, for example, the phone number or unique information of a mobile phone when the user uses a mobile phone . Further, the 'action type' includes various types of actions, such as view (action of viewing summary information about a document, previewing a document, etc.), play (action
of playing video, music, images, etc.), detailed view (action of viewing the full text of a document rather than summary information) , save (action of saving or storing a document) , buy (action of purchasing a document or content for sale) , recommend (action of recommending a document or content to another person) , evaluate (action of representing an individual's opinion about a document or content as digitized or standardized information) , attach supplementary information (action of attaching supplementary information, such as a comment or tag, to a document or content) , and bookmark
(action of storing the address of a document or content to view the document or content later, or a similar action) .
Meanwhile, the 'supplementary data' may include environmental information about the action taken by the user, for example, the location of the user, the environmental conditions (travel) , etc.
FIG. 2 is a diagram showing the detailed construction of the information search ranking system 200 according to an embodiment of the present invention. As shown in FIG. 2, the information search ranking system 200 includes a document collection module 210, a search engine 220, a UAL collection module 230, an attention value calculation module 240, a rank value calculation module 250, a document DB 10, a UAL DB 20, and an attention rank DB 30. The document collection module 210 collects a plurality of documents over the information network, and stores the documents in the document DB 10 so that the documents are
indexed and stored in order to respond to a keyword-based search request (a search request received from the document- search user terminal) .
The search engine 220 is an engine provided with functions required for a typical search operation of searching the document DB 10 for an input keyword and providing search results. In accordance with the features of the present invention, the search engine 220 provides ranked search results while referring to the Attention Ranks (ARs) of the attention rank DB 30, respectively related to the documents constituting the search results, assigns Rank Values (RVs) , in which the 'users' attention levels' are taken into account, to the documents, and provides the documents, to which the RVs are assigned, in the form of a ranked list. This procedure is described in detail below.
In the present invention, the 'user's attention level' is a principal factor for evaluating documents (content) , and is obtained by digitizing a user' s attention level using a memory model of the user. As is well known to those skilled in the art, a human being recognizes a certain fact or an object, and then gradually forgets the fact or the object as time elapses. This is the memory model of a human being, represented by a sigmoid function. In the prior art, in order to digitize the fact that users' attention is concentrated on a specific document, link information (the PageRank of Google) or the total number of hits for a given document is generally used as an index. However, such an index is only a simple
accumulative index, which does not take into account the temporal component of the memory model .
Meanwhile, as described above, the UAL collection module 230 collects User Action Logs (UALs) from the document-using user terminal 100 or the network provider server, and stores and manages the UALs in the UAL DB 20. The UALs can be collected from a logging tool (software organized in a web browser) installed in the document-using user terminal 100, or from UALs stored in the network provider server. It will be apparent that the two collection methods can be used together.
As shown in FIG. 3, the attention rank calculation module 240 includes the attention value calculation unit 242 and the action influence value calculation unit 244, thus calculating an Attention Rank (AR) from the multiplication of an Attention Value (AV) for all actions taken by each user by the Influence Value (IV) of the user for each action, with respect to all users who access an individual document (p of the following Equation [1] ) , and updating the calculated AR in the attention rank DB 30. The Attention Rank (AR) is defined by the following Equation [1] .
AR(p)= ∑[AV(ehpk)xIV(h)] [1] forall h,k
In Equation [1] , the AV, calculated by the attention value calculation unit 242, is represented by the following Equation [2] ,
where eh pk means that a user h takes an action I on a document
p , wk is a predetermined weight previously assigned to the type of action k , t is the difference between the time at which the AV is calculated and the time at which the action k actually occurs, that is, the elapsed time from the action k (hereinafter referred to as an 'elapsed time from an action' ) .
The weight has a value ranging from 0 to 1 according to the type of action. For example, for 'view' , a weight can be defined as 0.2, for 'play', as 0.4, for 'detailed view', as 0.5, and for 'recommend', as 0.9. Such an Equation [2] is a sigmoid function having the elapsed time from an action t as a variable, as defined in the right term thereof, and is implemented by modeling the memory of a human being, as described above. Therefore, the weight wk decreases as the elapsed time from the action increases.
The influence value (IV) of the user who took the action must be taken into account in the AV calculated by the attention value calculation unit 242. For example, if it is assumed that there is a user who took 1,000 actions and a user who took 10 actions, it is determined that the former user participates in the document more than the latter user by a factor of 100 times. In other words, when users' attention levels paid to the corresponding document are calculated, the user who took 1,000 actions has an excessively large influence on the document, compared to the remaining users. Therefore, a correction that allows the influence value (IV) of the user calculated by the action influence value calculation unit 244
to be taken into account in the AV must be performed. The influence value (IV) of the user for a given action is represented by the following Equation [3] ,
where σ is the sigmoid function, ch is the total number of all actions taken by the user h , m is the value obtained by dividing the sum of values ch by the total number of users
(m = for ΣaU«h. ). Such Equation [3] can be summarized as number of users
'standard deviation of the actions of the users' . Meanwhile, when the search engine 220 extracts documents matching the given keyword from the document DB 10, the rank value calculation module 250 calculates a Rank Value (RV) with reference to the AR for each extracted document, and thus ranks the documents. In this case, only the AR can be used as a factor for calculating the RV, but the degree of relation of each document to a keyword can be additionally calculated using the following Equation [4] according to the circumstances,
RV(p) = DocRel(p)xAR(p) [4] where DocRel is the degree of relation of a predetermined document to the keyword. For the calculation of the degree of relation of each document, a plurality of examples, implemented using various algorithms, is disclosed in the prior art, and thus a detailed description thereof is omitted
in the present invention.
Hereinafter, an information search ranking method according to an embodiment of the present invention is described with reference to FIG. 4. First, the document collection module 210 collects and stores a plurality of documents over the information network at step SlOO . This step is performed by a typical information search system at predetermined periods .
Then, the UAL collection module 230 collects UALs for documents from the document-using user terminal 100 or the network provider server, and stores the UALs at step S200.
The attention rank calculation module 240 calculates an AR in consideration of both the AV for all actions, taken by each user, and the IV of the user for each action, with respect to all users who took actions on each document (each document stored in the document DB) , on the basis of the UALs collected at step S200, at step S300.
In response to a keyword-based search request received from the document-search user terminal 300, the search engine 220 searches the document DB 10 for documents matching a keyword at step S400. The rank value calculation module 250 calculates the Rank Values (RVs) of found documents with reference to ARs for respective found documents stored in the attention rank DB 30, and ranks the found documents at step S500. At this step, as described above, the degree of relation of each document can be taken into account in the calculation of the RVs.
Next, the search engine 220 provides the ranked search results to the document-search user terminal 300 that requested the search at step S600. Through a series of these steps, a querying person (user) can view excellent search results, in which the 'user's attention levels' are taken into account .
[industrial Applicability]
As described above, the present invention provides an information search ranking system and method, which use actions of users taken on an individual document as users' attention levels based on the memory model of a human being, thus providing excellent ranked search results sensitive to the users' attention or preference in response to a keyword- based search request. Further, the present invention can provide excellent ranked search results, even for the recent rapid proliferation of User Created Content (UCC) .
Although the preferred embodiments of the present invention have been disclosed for illustrative purposes, those skilled in the art will appreciate that various modifications, additions and substitutions are possible, without departing from the scope and spirit of the invention as disclosed in the accompanying claims .
Claims
[CLAIMS]
[Claim l]
An information search ranking system based on users' attention levels, the ranking system including a search engine for searching documents stored in a document Database (DB) in response to a search request, and providing ranked search results, comprising: a User Action Log (UAL) collection module for collecting a plurality of UALs from at least one user terminal or at least one network provider server connected to an information network; an attention rank calculation module for calculating an Attention Rank (AR) from a multiplication of an Attention Value (AV) for all actions taken by each user by an Influence Value (IV) of the user for each action, with respect to all users who access an individual document stored in the document DB, on a basis of the collected UALs, and updating the calculated AR in an attention rank DB; and a rank value calculation module for calculating a Rank Value (RV) with reference to the AR stored in the attention rank DB, with respect to each document searched by the search engine in the document DB in response to a keyword-based search request received from the user terminal, thus ranking the documents .
[Claim 2]
The information search ranking system according to claim
1, further comprising a document collection module for collecting a plurality of documents over the information network and indexing and storing the collected documents in the document DB.
[Claim 3]
The information search ranking system according to claim 1, wherein the attention rank calculation module comprises an attention value calculation unit for calculating the AV by multiplying a predetermined weight ( wk ) , assigned to each action taken by a certain user accessing the individual document, by a sigmoid function that uses an elapsed time from the action (t) as a variable.
[Claim 4]
The information search ranking system according to claim 1 or 3 , wherein the weight ( wk ) has a value ranging from 0 to 1.
[Claim 5]
The information search ranking system according to claim
1, wherein the attention rank calculation module comprises an action influence value calculation unit for calculating the
Influence Value (IV) , the IV being calculated by the following equation:
where σ is a sigmoid function, ch is a total number of all actions taken by a user h , and m is a value obtained by- dividing a sum of values ch by a total number of users.
[Claim 6] The information search ranking system according to claim 1, wherein the rank value calculation module calculates the rank value by allowing a degree of relation of each document to the keyword to be taken into account in the attention rank.
[Claim 7] The information search ranking system according to claim 1, wherein the User Action Logs (UALs) comprise a document identifier, an action identifier, an action type, and an action time.
[Claim 8] An information search ranking method based on users' attention levels, the method providing ranked search results, comprising: a first step of collecting a plurality of documents over an information network and storing the documents in a document DB; a second step of collecting and storing User Action Logs (UALs) from at least one user terminal or at least one network provider server; a third step of calculating an Attention Rank (AR) from
a multiplication of an Attention Value (AV) for all actions taken by each user by an Influence Value (IV) of the user for each action, with respect to all users who access an individual stored document, on a basis of the collected UALs, and storing the calculated AR in an attention rank DB; and a fourth step of calculating a Rank Value (RV) with reference to the AR stored in the attention rank DB, with respect to each document searched in the document DB in response to a keyword-based search request received from the user terminal, thus providing ranked search results.
[Claim 9]
The information search ranking method according to claim 8, wherein the Attention Value (AV) at the third step is calculated by multiplying a predetermined weight ( wk ) , assigned to each action taken by a certain user accessing the individual document, by a sigmoid function that uses an elapsed time from the action (t) as a variable.
[Claim 10]
The information search ranking method according to claim 8, wherein the Influence Value (IV) is calculated by the following equation:
where σ is a sigmoid function, ch is a total number of all actions taken by a user h , and m is a value obtained by
dividing a sum of values ch by a total number of users .
[Claim ll]
The information search ranking method according to claim 8, wherein the Rank Value (RV) at the fourth step is calculated by allowing a degree of relation of each document to the keyword to be taken into account in the attention rank.
[Claim 12]
The information search ranking method according to claim 8, wherein the User Action Logs (UALs) comprise a document identifier, an action identifier, an action type, and an action time.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
KR10-2007-0041741 | 2007-04-30 | ||
KR1020070041741A KR100923505B1 (en) | 2007-04-30 | 2007-04-30 | Ranking system based on user's attention and the method thereof |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2008133368A1 true WO2008133368A1 (en) | 2008-11-06 |
Family
ID=39925803
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/KR2007/002701 WO2008133368A1 (en) | 2007-04-30 | 2007-06-04 | Information search ranking system and method based on users' attention levels |
Country Status (2)
Country | Link |
---|---|
KR (1) | KR100923505B1 (en) |
WO (1) | WO2008133368A1 (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2011139491A2 (en) * | 2010-04-30 | 2011-11-10 | Microsoft Corporation | Prioritization of resources based on user activities |
US8819009B2 (en) | 2011-05-12 | 2014-08-26 | Microsoft Corporation | Automatic social graph calculation |
US8979538B2 (en) | 2009-06-26 | 2015-03-17 | Microsoft Technology Licensing, Llc | Using game play elements to motivate learning |
US9477574B2 (en) | 2011-05-12 | 2016-10-25 | Microsoft Technology Licensing, Llc | Collection of intranet activity data |
US9697500B2 (en) | 2010-05-04 | 2017-07-04 | Microsoft Technology Licensing, Llc | Presentation of information describing user activities with regard to resources |
WO2022187478A1 (en) * | 2021-03-03 | 2022-09-09 | Home Depot International, Inc. | Attentive pseudo-relevance feedback network for query categorization |
US11960555B2 (en) | 2022-03-02 | 2024-04-16 | Home Depot Product Authority, Llc | Attentive pseudo-relevance feedback network for query categorization |
Families Citing this family (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR101078864B1 (en) | 2009-03-26 | 2011-11-02 | 한국과학기술원 | The query/document topic category transition analysis system and method and the query expansion based information retrieval system and method |
KR101218141B1 (en) * | 2011-09-27 | 2013-01-03 | (주)레드테이블 | Method and system for calculating ranking |
KR101894419B1 (en) * | 2011-12-12 | 2018-09-04 | 에스케이플래닛 주식회사 | System for providing personalized information, method thereof, and recordable medium storing the method |
KR101624284B1 (en) * | 2014-08-06 | 2016-06-08 | 네이버 주식회사 | System and method for providing information |
KR101649146B1 (en) | 2015-01-15 | 2016-08-19 | 주식회사 카카오 | Method and server for searching |
KR101593876B1 (en) * | 2015-04-27 | 2016-02-16 | 연세대학교 산학협력단 | Method and system for assessing environmental impacts capable of establishing weighting factors for environmental impact categories using internet search statistics |
CN105243124B (en) * | 2015-09-29 | 2018-11-09 | 百度在线网络技术(北京)有限公司 | Combination of resources processing method and processing device |
KR102008386B1 (en) * | 2017-09-14 | 2019-08-08 | 인하대학교 산학협력단 | Patent search engine evaluation system based on recall and method thereof |
KR102008387B1 (en) * | 2018-04-30 | 2019-08-07 | 인하대학교 산학협력단 | Patent search engine evaluation system based on non-recall and method thereof |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20020043679A (en) * | 2000-12-02 | 2002-06-12 | 김동현 | System and method for supporting search continuously according to user's inclination, and storage media having program source thereof |
US6665655B1 (en) * | 2000-04-14 | 2003-12-16 | Rightnow Technologies, Inc. | Implicit rating of retrieved information in an information search system |
KR20040006515A (en) * | 2002-07-12 | 2004-01-24 | 주식회사 네오위즈 | Method And System for Providing Information Service System and Searching Result by Using Log Analysis and Information Inputed by User |
KR20050095230A (en) * | 2004-03-25 | 2005-09-29 | 주식회사 첫눈 | Method and system for providing information service and information search service by using visited uniform resource locator log |
KR20060050397A (en) * | 2004-10-05 | 2006-05-19 | 마이크로소프트 코포레이션 | Systems, methods, and interfaces for providing personalized search and information access |
-
2007
- 2007-04-30 KR KR1020070041741A patent/KR100923505B1/en active IP Right Grant
- 2007-06-04 WO PCT/KR2007/002701 patent/WO2008133368A1/en active Application Filing
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6665655B1 (en) * | 2000-04-14 | 2003-12-16 | Rightnow Technologies, Inc. | Implicit rating of retrieved information in an information search system |
KR20020043679A (en) * | 2000-12-02 | 2002-06-12 | 김동현 | System and method for supporting search continuously according to user's inclination, and storage media having program source thereof |
KR20040006515A (en) * | 2002-07-12 | 2004-01-24 | 주식회사 네오위즈 | Method And System for Providing Information Service System and Searching Result by Using Log Analysis and Information Inputed by User |
KR20050095230A (en) * | 2004-03-25 | 2005-09-29 | 주식회사 첫눈 | Method and system for providing information service and information search service by using visited uniform resource locator log |
KR20060050397A (en) * | 2004-10-05 | 2006-05-19 | 마이크로소프트 코포레이션 | Systems, methods, and interfaces for providing personalized search and information access |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8979538B2 (en) | 2009-06-26 | 2015-03-17 | Microsoft Technology Licensing, Llc | Using game play elements to motivate learning |
WO2011139491A2 (en) * | 2010-04-30 | 2011-11-10 | Microsoft Corporation | Prioritization of resources based on user activities |
WO2011139491A3 (en) * | 2010-04-30 | 2012-02-09 | Microsoft Corporation | Prioritization of resources based on user activities |
CN102870113A (en) * | 2010-04-30 | 2013-01-09 | 微软公司 | Prioritization of resources based on user activities |
JP2013529332A (en) * | 2010-04-30 | 2013-07-18 | マイクロソフト コーポレーション | Prioritize resources based on user activity |
US9697500B2 (en) | 2010-05-04 | 2017-07-04 | Microsoft Technology Licensing, Llc | Presentation of information describing user activities with regard to resources |
US8819009B2 (en) | 2011-05-12 | 2014-08-26 | Microsoft Corporation | Automatic social graph calculation |
US9477574B2 (en) | 2011-05-12 | 2016-10-25 | Microsoft Technology Licensing, Llc | Collection of intranet activity data |
WO2022187478A1 (en) * | 2021-03-03 | 2022-09-09 | Home Depot International, Inc. | Attentive pseudo-relevance feedback network for query categorization |
US11960555B2 (en) | 2022-03-02 | 2024-04-16 | Home Depot Product Authority, Llc | Attentive pseudo-relevance feedback network for query categorization |
Also Published As
Publication number | Publication date |
---|---|
KR100923505B1 (en) | 2009-11-02 |
KR20080096887A (en) | 2008-11-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2008133368A1 (en) | Information search ranking system and method based on users' attention levels | |
US8818995B1 (en) | Search result ranking based on trust | |
US8589373B2 (en) | System and method for improved searching on the internet or similar networks and especially improved MetaNews and/or improved automatically generated newspapers | |
KR101171405B1 (en) | Personalization of placed content ordering in search results | |
US9094478B2 (en) | Prereading method and system for web browser | |
US7987261B2 (en) | Traffic predictor for network-accessible information modules | |
AU2011201819B2 (en) | Propagating useful information among related web pages, such as web pages of a website | |
USRE48437E1 (en) | Collecting and scoring online references | |
US8645367B1 (en) | Predicting data for document attributes based on aggregated data for repeated URL patterns | |
US20060064411A1 (en) | Search engine using user intent | |
US20120089598A1 (en) | Generating Website Profiles Based on Queries from Websites and User Activities on the Search Results | |
US20110246485A1 (en) | Systems and methods for grouping users based on metadata tag relevance ratings | |
US20110191331A1 (en) | Method of and System for Enhanced Local-Device Content Discovery | |
US7216122B2 (en) | Information processing device and method, recording medium, and program | |
CN101568921A (en) | Dynamic pricing models for digital content | |
US20100293448A1 (en) | Centralized website local content customization | |
CN102037464A (en) | Search results with most clicked next objects | |
US20180032614A1 (en) | System And Method For Compiling Search Results Using Information Regarding Length Of Time Users Spend Interacting With Individual Search Results | |
US9275145B2 (en) | Electronic document retrieval system with links to external documents | |
JP4875911B2 (en) | Content identification method and apparatus | |
Wang et al. | A personalization-oriented academic literature recommendation method | |
CN101788981A (en) | Deep web mobile search method, server and system | |
KR100371805B1 (en) | Method and system for providing related web sites for the current visitting of client | |
CN115048581B (en) | Network data analysis processing method, system and computer storage medium | |
CN113259150B (en) | Data processing method, system and storage medium |
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: 07746831 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 07746831 Country of ref document: EP Kind code of ref document: A1 |