JP2013122701A - Information retrieval device and information retrieval method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To perform suitable recommendation according to an operation history of a user.SOLUTION: The information retrieval device includes profile generation means for generating profile information to be used for recommendation from document operation history information recorded with an operation history for each operation type of a user performing in relation to a plurality of registered documents, recommendation score calculation means for analyzing the profile information generated by the profile generation means and calculating each recommendation score of the plurality of documents on the basis of the analysis result, and recommendation result display means for displaying a recommendation result document calculated by the recommendation score calculation means on a display device, to solve a problem in which many unnecessary documents appear in the recommendation result and the suitable recommendation cannot be performed.

Description

  The present invention relates to an information search apparatus and an information search method, and more particularly, to a technique suitable for recommending a related document when a user views the document.

  As a conventional search system, a type of searching for content that matches a keyword input by a user has been mainstream. This is because if a correct keyword is input, a search result can be obtained at high speed, but the user needs to consider a new keyword every time a search is performed. In addition, there is a problem that it is difficult for a general unskilled user to construct a correct keyword.

  In order to solve such a problem, a technique has been proposed in which a user's preference is estimated from a history of using the content, and information recommendation (recommendation) for presenting the content estimated to be preferred by the user is performed. .

  Patent Document 1 is a technique for presenting a book separately purchased by a person who has purchased the book to be referred to based on a past user purchase history when the user references the book on an Internet book sales site. In Patent Document 1, a history of contents purchased by a plurality of users in the past is statistically learned from a large amount of purchase histories, and contents are recommended to users having similar preferences that are patterned without specifying an individual. In an EC (electronic commerce) site on the Internet, such as Patent Document 1, the focus is on one index “purchase” in order to estimate a condition that a user likes the content.

  On the other hand, in a document management system in an enterprise environment, a user can perform various operations on a document, and each operation is not an index indicating that the user likes content more strongly than “purchase”. Therefore, in an office document management system, it is necessary to estimate the relationship between a user and a document, and between a document and a document, not from a single operation but from each of a plurality of operations.

  Patent Document 2 is a technique for presenting a document related to such an office environment. Patent Document 2 manages access to a set of documents from creation to end of a document as a case session, links related documents based on the operation content of each case session, and a document is viewed When it is done, the related document is displayed based on the related link.

US Pat. No. 6,266,649 JP 2011-28447 A

  The related document display system proposed in Patent Document 2 uses document creation start information as a recommendation, but does not consider what intention the creation was made based on. In addition, a link between documents is created by a copy-and-paste operation from document to document, but a relationship between documents is simply added.

For example, when creating an office document, there are various differences in the purpose of creation. For this reason, as in Patent Document 2, if all these creation operations are handled in the same way, many unnecessary documents appear in the recommendation result, and there is a problem that a suitable recommendation cannot be performed.
The present invention has been made in view of the above-described problems, and it is an object of the present invention to make a suitable recommendation according to a user's operation history.

  The information search apparatus according to the present invention is a profile generation unit that generates profile information used for recommendation from document operation history information in which operation history for each operation type of a user performed on a plurality of registered documents is recorded. A recommendation score calculation means for analyzing the profile information generated by the profile generation means and calculating a recommendation score for each of the plurality of documents based on the analysis result, and a recommendation result document calculated by the recommendation score calculation means And a recommendation result display means for displaying on a display device.

  According to the present invention, unnecessary recommended documents can be removed, and the accuracy of document search and document recommendation can be improved.

It is a block diagram which shows the structural example of the information search device of 1st Embodiment. It is a block diagram showing the main components of a document management system and the flow of data. It is a figure which shows the example of the operation log of 1st Embodiment. It is a figure which shows the example of the user interface for implement | achieving (function 1). It is a figure which shows the example of the user interface for implement | achieving (function 2). It is a flowchart which shows the flow of a process of a profile production part. It is a figure which shows the structural example of a document relevance degree table. It is a figure which shows the structural example of a document relevance degree table. It is a figure which shows the structural example of a copy classification table. It is a figure which shows the example of the copy operation classification | category of 1st Embodiment. It is a flowchart explaining the flow of a process which classifies a copy log in detail. It is a figure which shows the example of the copy operation classification | category of 1st Embodiment. It is a flowchart which shows the flow of the recommendation process by a document relevance degree. It is a flowchart which shows the flow of the process of relevance calculation by simultaneous use. It is a figure which shows an example of the document relevance degree application strategy of 1st Embodiment. It is a flowchart which shows the flow of a relevance score correction process. It is a flowchart which shows the flow of a process of a document importance calculation part. It is a figure which shows the example of the application strategy of the document importance calculation of 1st Embodiment. It is a flowchart which shows the flow of the process which correct | amends the importance score of a document.

(First embodiment)
Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings.
FIG. 1 is a block diagram illustrating a schematic configuration of an information search apparatus 100 according to the present embodiment. In FIG. 1, the CPU 101 controls the entire apparatus by executing a control program stored in the ROM 102.

The ROM 102 stores programs executed by the CPU 101 and various parameters. The program is executed by the CPU 101 to cause the information search apparatus 100 of the present embodiment to function as various means for executing each process shown in the flowchart described below.
The RAM 103 stores a document to be processed and various data. Further, it is used for developing a program to be executed by the CPU 101.

The storage device 104 is a device that stores document data and operation log information to be processed by the information search device 100 of the present embodiment.
The display 105 displays a user interface for operating the processing of this embodiment. When a document recommendation command is issued by operating the user interface displayed on the display 105, a recommended document calculation process is performed by the processing program expanded in the RAM 103. The processing result is stored again in the storage device 104 such as a hard disk and can be confirmed on the user interface displayed on the display 105.
In this embodiment, processing corresponding to each step of the flowchart described later is realized by software using a computer (CPU 101), but part or all of the processing is realized by hardware such as an electronic circuit. It may be.

FIG. 2 is a block diagram showing the main components of the document management system 200 configured by the CPU 101 of the information search apparatus 100 according to the present embodiment executing a program stored in the ROM 102 and the flow of data.
The document management system 200 detects an operation type for a large number of registered documents and records it in the storage device 104 as operation log information. The detected operation types include, for example, document registration, document open, document close, browsing, editing, copying, and moving.

  An operation log 201 stores document operation history information detected by the information search apparatus 100. The operation log 201 stores information such as the operation date and time, the operation user, and the operation target document in addition to the operation type described above. Details of the operation log 201 will be described later.

  A profile generation unit 202 analyzes the operation log 201 and generates profile information 203 of a user or a document used for recommendation (information recommendation). The profile information 203 stores user and document analysis information extracted from the log, such as browsing of each document, the number of times of printing, and the number of times each user operates each document. Details of the profile information 203 will be described later.

  A recommendation score calculation unit 204 calculates a recommendation score for each document based on the analysis result of the profile information 203, and includes a document importance calculation unit 205 and a document relevance calculation unit 206. There are two types of recommendation scores described in the recommendation score calculation unit 204 of the present embodiment. One is a score indicating the importance of each document, and is calculated by the document importance calculator 205. The other is a score indicating the degree of association between documents, and is calculated by the document association degree calculation unit 206. Reference numeral 207 denotes a recommendation result display unit that displays a recommendation result document based on the recommendation score given by the recommendation score calculation unit 204.

[Operation log]
FIG. 3 shows details of the operation log 201 of FIG.
FIG. 3A is a diagram illustrating an example of the structure of information stored in the operation log 201. The operation log 201 stores at least five pieces of information from 301 to 305 for each operation.
The first information 301 is the time when the operation was performed. Here, the local time held by the OS (operating system) is used.

The second information 302 is an operation user name or an operation user identifier. Before operating the document, the user performs a login operation to the information search apparatus 100 and notifies the user information in advance.
The third information 303 is an operation target document name or identifier.
The fourth information 304 is an operation type. The operation type stores the type of operation on the document detected on the user interface of the information search apparatus 100. Here, “check in”, “check out”, “login”, “logout”, “document registration”, “document open”, “document close”, “document edit”, “document copy”, “document move” , “Document printing” is stored.
The fifth information 305 stores a copy destination and a destination file name (new file name) when the “document copy” and “document move” operations are performed.

FIG. 3B is a diagram illustrating an example of a user operation restored from the log stored in FIG.
From the log in FIG. 3A, User X copies the last week's weekly report “Weekly_20110302.doc” and renames it to “Weekly_20110309.doc” including the date of this week. Furthermore, a series of operations in which User A and User B have updated “Weekly_20110309.doc” are restored.

[Recommendation result display]
4 and 5 are diagrams showing specific examples displayed on the recommendation result display unit 207 of the document management system 200 of FIG. The recommendation function of this embodiment has the following two.
(A) A function (function 1) that recommends a document related to the selected document when the user selects a document on the document management system.
(B) A function (function 2) that, when the user selects a folder on the document management system, important documents existing in the selected folder and its subfolders are recommended.

4 shows a specific example of (Function 1), and FIG. 5 shows a specific example of (Function 2).
Reference numeral 400 in FIG. 4 denotes a user interface of the document management system for realizing (Function 1).
In FIG. 4, reference numeral 401 denotes a folder view, which displays all folders registered in the document management system 200 in a tree format.
Reference numeral 402 denotes a file view. The file view 402 displays a document registered in the document management system 200 and a file that exists immediately under the folder selected in the folder view 401.

  Reference numerals 403 and 404 denote specific examples displayed on the recommendation result display unit 207. A recommendation result document display unit 403 recommends a related document of the document selected in the file view 402. In the example of FIG. 4, immediately after the document A is selected in the file view 402, the document X and the document Y that are related documents of the document A are recommended.

  Reference numeral 404 denotes a recommendation score display unit. The recommendation score is determined by inter-document relevance calculation processing described later, and the recommendation result document is sorted and presented by this score. In the example of FIG. 4, the degree of association between the document A and the document X is 95, the degree of association between the document A and the document Y is 60, and the degree of association of the other documents with respect to the document A is less than 60.

5 is a user interface of the document management system for realizing (Function 2).
A folder view 501 displays all folders registered in the document management system 200 in a tree format.
Reference numeral 502 denotes a file view. In 502, a file registered in the document management system 200 and a file existing immediately below the folder selected in the folder view 501 is displayed. In this example, a document A, a document B, a document C, and a document D exist immediately below the folder T.

Reference numerals 503 and 504 denote specific examples displayed on the recommendation result display unit 207. Reference numeral 503 denotes a recommendation result document display unit that recommends a document having a high importance, which will be described later, from all files existing under the folder selected in the folder view 501 and its subfolders. In this example, immediately after the folder T is selected in the folder view 501, the document K and the document L, which are important documents existing under the folder T, are recommended.
Reference numeral 504 denotes a recommendation score display unit. The recommendation score is determined by an inter-document importance calculation process described later, and the recommendation result document is sorted and presented by this score. In the example of FIG. 5, the importance level of the document K is 88, the importance level of the document L is 50, and the importance levels of other documents existing in the folder T and below are less than 50.

[Create Profile]
FIG. 6 is a flowchart showing a flow of processing performed by the profile generation unit 202 of FIG. The profile information 203 includes a document relevance table for storing document relevance information, a document importance table for storing document importance information, and a copy classification table for correcting a recommendation score according to a copy operation classification result. The profile generation unit 202 updates the information stored in these three tables. The processing corresponding to each step in the flowchart of FIG. 6 is realized by the CPU 101 developing the program recorded in the ROM 102 on the RAM 103 functioning as a work memory and executing it.

In step S <b> 601, the first log record is acquired from the operation log 201.
In step S602, the information in the document relevance table corresponding to the log record acquired in step S601 is updated. Details of the document relevance table will be described later.
In step S603, the information in the document importance table corresponding to the log record acquired in step S601 is updated. Details of the document importance table will be described later.

In step S604, the copy classification table information corresponding to the log record acquired in step S601 is updated. Details of the copy classification table will be described later.
In step S605, it is determined whether all log records of the operation log 201 have been checked. If there is a record that has not been checked yet, the process proceeds to S606, the next log record is acquired, and then the process proceeds to S602. Exit if all log records have been examined. The records of the operation log 201 increase every moment according to daily user operations. In order to reflect the latest operation log information in the recommendation result, it is desirable to periodically execute this profile generation process once every several hours or once a day.

FIG. 7 is a diagram showing a configuration example of the document relevance table updated in S602 of FIG.
700 in FIG. 7 shows an example of a document relevance table. As shown in FIG. 7, all the documents registered in the document management system 200 are stored in the row 701 and the column 702 of the document relevance degree table 700, and relevance is given to all the documents. The numbers at the intersections of the documents in row 701 and column 702 indicate the degree of association between documents.

  In this embodiment, it is assumed that “documents used simultaneously” are related documents. Documents used at the same time refer to documents that a user browses and prints at the same time. For example, if logs of document 1 open, document 2 open, document 1 close, and document 2 close are obtained from the operation log 201 in chronological order, there is a period in which document 1 and document 2 are opened simultaneously. In such a case, it is determined that the documents 1 and 2 are used at the same time, and +1 is added to the numerical value of the intersection of the documents 1 and 2 in the document relevance table 700. In this way, the number of times of simultaneous use between all the documents in the document relevance table 700 is accumulated.

  Here, in the present embodiment, the case where documents are opened at the same time has been described. At the same time, documents that are checked in / out, logged in / out, and documents that are printed during these operations can also be used as documents that are used at the same time. Further, the standard for assigning the document relevance is not limited to documents used at the same time. Document relevance can be adaptively given by a copy source document (replication source document) / copy destination document (replication destination document), a document linked by a hyperlink, a document having similar contents, or the like.

800 in FIG. 8 is a diagram showing a configuration example of the document importance level table updated in S603 in FIG. In the present embodiment, it is assumed that a document that has been viewed (opened) and printed by a plurality of users is a highly important document.
A row 801 of the document importance level table 800 indicates all documents registered in the document management system 200, and a column 802 indicates an open operation and a print operation. The numerical value at the intersection of row 801 and column 802 indicates the number of times the document has been manipulated. In FIG. 8, document 1 is viewed 10 times and printed 5 times.

900 in FIG. 9 is a diagram illustrating a configuration example of the copy classification table updated in S604 in FIG.
Reference numeral 901 denotes a copy source document in which a copy source document in the past is extracted from the operation log 201 and stored. Reference numeral 902 denotes a copy destination document list corresponding to the copy source document 901. The structure of the copy destination document list 902 stores a pair record of {copy destination document, classification number}. When a log in which a plurality of copies are made from one document is detected, the element of the copy destination document list 902 has a plurality of records. The classification that is an element of the copy destination document list 902 is a classification based on an estimation of the purpose of the document creation by the user from the log record of the copy operation. Details of the classification are shown in FIG.

FIG. 10 is a diagram for explaining the classification numbers of the copy destination document list 902 in FIG.
In the present embodiment, the purpose of the user creating the document is estimated from the copy operation log extracted from the operation log 201, and the document is classified according to the estimation result. In a document importance level calculation process and a document relevance level calculation process, which will be described later, a suitable recommendation score corresponding to the classification is given.
The contents of the classification are six patterns of “difference update”, “backup”, “template”, “document division (1)”, “document division (2)”, and “partial extraction”.

“Differential update” means that, for example, an old version (old version) document containing a past date is copied (duplicated), the file name is changed to the current date, and the file on the current date is updated. It is a case.
“Backup” is a case where, for example, a certain file is copied and renamed to a file name including “_old” and stored.

The “template” is a case where the copy destination document is updated using the design of the copy source document as it is.
“Document division (1)” is a case where one document is copied and divided into two documents, and the contents of the copy source document are represented by two documents.
“Document division (2)” is a case where a plurality of copies are created from one document, and the contents of the copy source document are represented by a plurality of copy destination documents without updating the copy source document.
“Partial extraction” is a case where the copy destination document is updated while leaving a part of the copy source document.

  The log patterns of the copy operation extracted from the operation log 201 are classified into these six patterns. As the classification criteria, the method described in the column “Determination method from log” in FIG. 10 is used. The processing flow of this determination method is shown in FIG. In the present embodiment, the patterns are classified into six patterns, but the present invention is not limited to this. As will be described later, it may be changed according to the score correction strategy by copy classification.

FIG. 11 is a flowchart for explaining the flow of processing for classifying the copy log of the operation log 201 in detail as shown in FIG. The processing corresponding to each step in the flowchart of FIG. 11 is realized by the CPU 101 executing the program recorded in the ROM 102 on the RAM 103 functioning as a work memory.
In step S1101, a log record whose “operation type” is “copy” is extracted from the operation log 201 and acquired. In the subsequent flowcharts, relevance determination processing is performed to determine to which classification in FIG. 10 the acquired copy operation belongs.

  In step S1102, it is determined whether both the copy destination document and the copy source document have been updated in the log in the vicinity of the acquired copy log. Here, as the nearby log, a log whose difference in recording time between the copy log and the update log is smaller than a certain threshold is used. For example, a case where both have been updated within 30 minutes is extracted. If both have been updated, it is determined in S1120 that the pattern is “document division (1)” in FIG. 10, and the process ends. If at least one of them has not been updated, the process proceeds to S1103.

  In S1103, it is determined whether there are a plurality of copy destination documents. If there is a copy log having the same copy source document in the vicinity of the acquired copy log, it is determined that there are a plurality of copy destination documents, and the process advances to step S1107 to determine whether all the copy destination documents have been updated. If not all have been updated, the process proceeds to S1104. If all have been updated, it is determined in S1121 that the pattern is “document division (2)” in FIG. 10, and the process ends.

  On the other hand, if the result of determination in S1103 is that there is no such event, the process proceeds to S1104. In step S1104, it is determined whether the copy destination document has been updated. If it has been updated, the process proceeds to S1105. If it has not been updated, the process proceeds to S1122, where it is determined that the pattern is the “backup” pattern in FIG. 10, and the process ends.

  In step S <b> 1105, it is determined whether the reduction rate of the document size before and after the update log operation is smaller than a certain threshold. If it is smaller, the process proceeds to S1106. If it is not smaller, it is determined that the content has been greatly cut after copying, and it is determined to be the “partial extraction” pattern of FIG. 10 (S1123).

  In step S1106, it is determined whether the copy source file is a copy source in many copy logs. Here, not only the vicinity of the copy log acquired in S1101, but also all the copy logs of the operation log 201 are scanned and determined. If it is the copy source in many logs, it is determined as the “template” pattern of FIG. 10 (S1124). If it is not the copy source, it is determined that the pattern is the “difference update” pattern (S1125).

FIG. 12 is a diagram illustrating an example of a neighborhood pattern of a copy log obtained from the operation log 201. How the copy log of FIG. 12 is classified into the classification items of FIG. 10 will be described with reference to the flowchart of FIG.
FIG. 12B is a diagram in which the operation indicated by the log record in FIG.
As shown in FIG. 12B, the log 1200 shows a copy of the document “patent specification.doc”, renames the file “patent specification_old.doc”, and saves it. Thereafter, a series of operations in which “patent specification.doc” of the copy source file is updated.

In S1102 of FIG. 11, the log 1200 is updated only for the copy source, and thus the process proceeds to S1103.
In S1103, since the log 1200 has one copy destination document, the process proceeds to S1104.
In S1104, since the copy destination document has not been updated in the log 1200, the process advances to S1122, and is classified as the “backup” pattern in FIG.
In this way, it is determined to which classification the copy operation belongs, and the classification number obtained from FIG. 10 is assigned to the copy destination document list 902 of FIG.
By performing the above processing, the profile generation unit 202 in FIG. 2 generates and stores information to be stored in the document relevance table 700, the document importance table 800, and the copy classification table 900 constituting the profile information 203 in FIG. To do.

[recommend]
Next, the recommendation score calculation unit 204 in FIG. 2 will be described.
In this embodiment, the document is recommended based on two criteria as described above. One is a recommendation based on the relationship between documents (function 1), and the other is a recommendation based on the importance of each document (function 2). The user operation using each function has been described with reference to FIGS.

[Recommendations based on document relevance]
FIG. 13 is a flowchart showing a flow of processing of a recommendation (function 1) based on the document relevance level. The processing corresponding to each step in the flowchart of FIG. 13 is realized by the CPU 101 executing the program recorded in the ROM 102 on the RAM 103 that functions as a work memory.
In S1301, the document selected by the user on the user interface in FIG. 4 is acquired.
In S1302, the relevance score is calculated and corrected so that the score of the document used simultaneously with the selected document acquired in S1301 is higher. Details of this step will be described later.
In S1303, the relevance score obtained in S1302 is corrected using correction terms specific to each copy classification shown in FIG. Details of this step will be described later.
In S1304, the documents scored in S1302 and 1303 are sorted and output in the order of score.

  FIG. 14 is a flowchart showing a flow of relevance calculation processing by simultaneous use performed in S1302 of FIG. In the present embodiment, the degree of relevance due to simultaneous use uses a value stored in the document relevance degree table 700 of FIG. 7 as a score. Since the value stored in the document relevance level table 700 indicates the frequency with which the document and the document are used simultaneously, the relevance score between documents increases as the frequency increases.

  In S1401, a record corresponding to the document acquired in S1301 of FIG. 13 is acquired from the document relevance table 700 of FIG. If the document selected in S1301 is “Document 1”, the first record in FIG. 7 is acquired.

In S1402, each column of the records acquired in S1401 is acquired as a scoring target document. If the record is “document 1”, the records are sequentially acquired from “document 2” in column 702.
In S1403, a value corresponding to the column 702 acquired in S1402 is acquired. In FIG. 7, the value 5 of the intersection of “Document 1” and “Document 2” is acquired.
In S1404, it is determined whether all the columns of the record acquired in S1401 have been examined. If not all are checked, the process proceeds to S1406, and the value of the next column is acquired. If all columns have been examined, the process proceeds to S1405.
In step S1405, a process for normalizing the relevance level of each document acquired from the document relevance level table 700 is performed. The score Normalized_score (i) of the document i is normalized by the maximum score as follows.

The relevance score correction by copy classification performed in S1303 of FIG. 13 will be described with reference to FIGS.
In this embodiment, the document creation operation is classified from the document copy log as shown in FIG. 15, and a recommendation according to each classification is performed.

FIG. 15 is a diagram illustrating an example of a document relevance application strategy for each classification.
There are two viewpoints corresponding to each column in FIG. 15 in obtaining the degree of association between documents from the copy operation. One is whether or not the related document for the copy source document is set as the related document for the copy destination document (related document takeover). The second point of view is how to assign the degree of association from the copy source to the copy destination or from the copy destination to the copy source (copy destination document relevance correction).

  As shown in FIG. 15, only the “difference update” pattern and the “template” pattern are not inherited for the related document of the copy source document. Other patterns are taken over. As a result, in the “difference update” pattern, it is possible to prevent a large amount of old versions of documents and related documents from being copied from being output as recommendation results. Also, in the “template” pattern, the copy source document that is a template is less likely to be recommended.

  Regarding the relevance between the copy source and the copy destination, the relevance of the copy source document from the copy source document of the “backup” pattern and the relevance of the copy source document from the copy destination of the “template” and “document division (2)” patterns. to correct. In FIG. 15, “no change” means that the recommendation score based on the document relevance calculation by simultaneous use in S1302 of FIG. 13 is used as it is, and “decrease” is in the direction of reducing the recommendation score. It means to correct. In the example of FIG. 15, the degree of relevance from the updated document to the backup document, the document created by copying the template from the template document, and the recommendation score from the divided document to the original document when the document is divided into a plurality of documents are reduced. It becomes difficult to be recommended.

FIG. 16 is a flowchart showing the flow of relevance score correction processing by copy classification performed in S1303 of FIG. 13 based on the application strategy of FIG. The processing corresponding to each step in the flowchart of FIG. 16 is realized by the CPU 101 executing the program recorded in the ROM 102 on the RAM 103 that functions as a work memory.
In S1601, it is determined whether or not a log in which the document selected in S1301 in FIG. If it exists, the process proceeds to S1602, and if it does not exist, the document relevance score in S1302 of FIG. 13 is ended as the final recommendation score.

  In step S1602, it is determined whether the selected document determined to have a copy log in step S1601 is a copy source document or a copy destination document in the operation log 201. If it is a copy source document, the process proceeds to S1603, and if it is a copy destination document, the process proceeds to S1606.

In S1603, a copy destination document is acquired when the selected document is a copy source document.
In step S1604, it is determined whether the copy log of the selected document is classified as a backup pattern in the classification of FIG. For the determination, the classification number in the copy classification table 900 of FIG. 9 is used. If it is a backup pattern, the process proceeds to S1605. If it is not a backup pattern, the document relevance score in S1302 of FIG. 13 is ended as the final recommendation score.

  In step S1605, correction is made in the direction of decreasing the recommendation score of the copy destination document as in the backup pattern of FIG. The final score Final_score (i) is corrected as follows with respect to the score of the copy destination document i shown in (Formula 1).

In S1606, when it is determined in S1602 that the selected document is a copy destination document, a copy source document in the copy log is acquired.
In S1607, it is determined whether or not the copy log of the selected document is classified into the difference update pattern in the classification of FIG. If it is a difference update pattern, the document relevance score in S1302 of FIG. 13 is ended as the final recommendation score. If it is not a difference update pattern, the process proceeds to S1608.

In step S1608, it is determined whether the copy log is a template pattern. If it is a template pattern, the process proceeds to S1611. If it is not a template pattern, the process proceeds to S1609.
In step S1609, the related document score of the copy source document is calculated. The calculation method is as shown in S1302 of FIG. As shown in FIG. 15, the related document of the copy source document is taken over as the related document of the selected document.

In S1610, it is determined whether or not the copy log is a document division (2) pattern. If it is a document division (2) pattern, the process advances to step S1611; otherwise, the process ends. The recommendation score at the end is the document relevance level due to the simultaneous use of the selected document and the document relevance level due to the simultaneous use of the copy source document of the selected document.
In S1611, correction is made in the direction of lowering the recommendation score of the copy source document as in the template pattern and document division (2) pattern of FIG. The final score Final_score (i) is corrected as follows with respect to the score of the copy destination document i shown in (Formula 1).

In this way, by correcting the recommendation score according to the copy classification pattern estimated from the operation log 201, the possibility that an unnecessary document appears in the related document recommendation result is reduced.
In this embodiment, scoring is performed from the two viewpoints of taking over the related document to the copy source document and the degree of relevance between the copy source document and the copy destination document. The score is calculated using either one of the viewpoints. May be. When either one is used, the copy classification range may be changed. For example, when only relevance correction is performed, when both the copy destination and the copy source are “no change” in FIG. 15, only the copy destination from the copy source is reduced, or only the copy source is reduced from the copy destination , Three classifications can be used.

[Recommendations based on document importance]
Next, the document importance calculation unit 205 in FIG. 2 that implements the recommendation based on the document importance in FIG. 5 will be described.
FIG. 17 is a flowchart showing a processing flow of the document importance degree calculation unit 205. The processing corresponding to each step in the flowchart of FIG. 17 is realized by the CPU 101 executing the program recorded in the ROM 102 on the RAM 103 that functions as a work memory.
In S1701, the folder selected in 501 in FIG. 5 is acquired.
In S1702, a scoring target document is acquired.
In S1703, it is determined whether the scoring target document acquired in S1702 exists in the selected folder and subfolder acquired in S1701. If it exists in the folder, the process proceeds to S1704. If it does not exist, the process proceeds to S1710 to set score 0, and the process proceeds to S1707.
In S1704, an importance score is calculated. In the present embodiment, a document that is frequently printed or viewed by a plurality of users is defined as a document having a high importance. The importance score based on the browsing / printing frequency for the document i is obtained by the following equation.

  Here, the printing frequency Print_frequency and the browsing frequency Open_frequency are values obtained from the column 802 of the document importance table 800 of FIG. τ and υ are weights that determine which of the printing frequency and the browsing frequency is important. Here, it is assumed that the printing weight is twice the viewing weight (τ = 2, υ = 1).

In step S1705, it is determined whether the copy log of the document acquired in step S1702 exists in the operation log 201. If there is a copy log, the process proceeds to S1706, and if not, the process proceeds to S1707.
In S1706, the importance score of the document is corrected according to the copy classification of FIG. Details of this processing will be described later.

In step S1707, it is determined whether scoring has been performed for all documents. If it has been scored, the process advances to step S1708. If not, the process advances to step S1709 to acquire the next document to be scored, and the process returns to step S1703.
In step S1708, the documents are sorted so that the importance scores assigned in steps S1701 to S1707 are in ascending order and output as recommendation results.

Next, with reference to FIGS. 18 and 19, the importance score correction by copy classification performed in S1706 of FIG. 17 will be described.
In this embodiment, the document creation operation is classified from the document copy log as shown in FIG. 10, and an appropriate recommendation is performed according to each classification.

  FIG. 18 is a diagram illustrating an example of an application strategy of document importance calculation for each classification. As shown in FIG. 18, the importance of the copy source or copy destination document is corrected according to each classification. In the differential update pattern, the importance of the previous version document that is the copy source is lowered. In the backup pattern, the importance of the backup document as the copy destination is lowered. In document division (2), since the copy source document can be reproduced with a plurality of copy destination documents, the importance of the copy source document is lowered.

FIG. 19 is a flowchart showing the flow of importance score correction processing performed in S1706 of FIG. 17 based on the application strategy of FIG.
In S1901, it is determined whether or not the scoring target document for which the copy log is found in S1705 in FIG. 17 is a copy source document. If it is a copy source, the process proceeds to S1902, and if it is a copy destination, the process proceeds to S1904.

  In S1902, it is determined whether the copy log of the document to be scored is a difference update pattern or a document division (2) pattern. If it is a difference update or document division (2) pattern, the process advances to step S1903, and the score of the copy source document is corrected in a direction of reducing the importance of the copy source document by the application strategy of FIG. The final importance score Final_score is corrected as shown in the following (Expression 5) with respect to the importance score by printing / viewing of (Expression 4).

In S1904, it is determined whether or not the copy log of the document to be scored is a backup pattern. If it is a backup pattern, the process advances to step S1905; otherwise, the process ends.
In step S1905, the importance level of the copy destination document is corrected based on the backup pattern application strategy shown in FIG. The correction is calculated in the same manner as (Equation 5) described above.
As described above, by appropriately determining the importance of the copy source document or the copy destination document based on the document generation pattern by copying, the possibility that an unimportant document is recommended when selecting a folder is reduced.

  In the present embodiment, the two indexes of the document relevance level and the document importance level are used for the recommendation, but either one may be used. When either one is used, the copy classification range may be changed. For example, when only the document importance is used as the recommendation index, in FIG. 18, when both the copy source and copy destination documents have the “no change” importance, the copy source document importance only “decreases”, the copy In the case where only the importance level of the previous document is “reduced”, the following three classifications are used.

(Second Embodiment)
In the first embodiment, the recommendation score based on the printing / viewing frequency is first calculated when calculating the recommendation score, and the copy classification correction is performed as post-processing. However, the values in the document relevance table 700 in FIG. 7 and the document importance table 800 in FIG. 8 may be reduced by using the direct copy history at the time of profile generation. This is performed as follows after the copy classification process of FIG. 11 is executed at the time of profile generation. That is, based on the scoring strategies of FIGS. 15 and 18, a negative integer as a penalty term is added to the values of the document relevance table and the document importance table corresponding to the document whose score is determined to be “decreased” This can be realized by subtracting a positive integer.

  For example, if the penalty term is -5 and the document 1 and the document 2 are the copy source document and the copy destination document of the backup pattern, respectively, the relevance from the document 1 to the document 2 is lowered by the application strategy of FIG. A penalty term is added to the intersection value 5 of document 1 and document 2 in FIG.

At the time of recommendation, the values of the document relevance table and the document importance table that have been subjected to copy classification correction as described above are directly output as recommendation scores.
In this way, the same effect as that of the first embodiment can be obtained by performing the copy classification correction at the time of profile generation.

As described above in the embodiment of the present invention, the intent of creating a document is estimated from the history of the user's operation of the document in the past, particularly the history of document copy, and based on the estimation result Recommended to do. Thereby, the suitable recommendation according to a user's operation history can be performed.
In addition, by estimating how the user created the document from the copy log and recommending the document according to the document relevance and document importance according to the estimation result, unnecessary noise in the recommendation result is reduced, and the recommendation is recommended. Accuracy can be improved.
Although the document recommendation function when the user selects a file or folder has been described in this embodiment, the present invention is not limited to these functions. The present invention can also be applied to a case where search results when a user searches for a document are sorted in order of document importance, or a case where related documents of a search result document are presented to the user in order of relevance.

(Other embodiments)
The present invention can also be realized by executing the following processing. That is, software (computer program) that realizes the functions of the above-described embodiments is supplied to a system or apparatus via a network or various computer-readable storage media. Then, the computer (or CPU, MPU, etc.) of the system or apparatus reads out and executes the program.

DESCRIPTION OF SYMBOLS 100 Information retrieval apparatus, 200 Document management system, 201 Operation log, 202 Profile production | generation part, 203 Profile information, 204 Recommendation score calculation part, 205 Document importance calculation part, 206 Document relevance calculation part, 207 Recommendation result display part

Claims (22)

Profile generation means for generating profile information to be used for recommendation from document operation history information in which operation history for each operation type of a user performed on a plurality of registered documents is recorded;
Analyzing the profile information generated by the profile generation unit, and calculating a recommendation score for each of the plurality of documents based on the analysis result; and
An information retrieval apparatus comprising: a recommendation result display means for displaying a recommendation result document calculated by the recommendation score calculation means on a display device. The profile information has a copy classification table for storing a copy source document and a copy destination document,
The profile generation means extracts a document that has become a copy source in the past from the document operation history information and stores it in the copy source document, and creates the document from the history indicating that the document has been copied. 2. The information retrieval apparatus according to claim 1, wherein the information is classified into predetermined classification items and stored in the duplication destination document.   3. The information search apparatus according to claim 2, wherein the classification items are classified by estimating what purpose the user has created a document from a log record of a copy operation. The profile information has a document relevance table in which all registered documents are stored and relevance between all documents is given,
The profile generation means updates the relevance information stored in the document relevance table every time the documents are used at the same time, and the relevance information is stored in the document relevance table. 4. The information search device according to 3.   4. The profile generation unit according to claim 2, wherein the profile generation unit determines whether to take over a document related to a copy source document as a document related to a copy destination document according to a result of classifying the document. Information retrieval device. The profile information has a document importance table storing importance information of all registered documents,
3. The profile generation unit assumes that a document viewed or printed by a user is a document having high importance, and updates the importance information of the document every time the user browses or prints. The information search device according to any one of? 4.   The classification items include a pattern for duplicating and updating an old version document, a pattern for duplicating an old version and storing it as a backup, a pattern for duplicating and updating a template, and a pattern for duplicating and dividing a document. The information search apparatus according to any one of claims 2 to 6, wherein the information search apparatus is characterized in that:   The information search apparatus according to claim 2, wherein the recommendation score calculation unit calculates an importance score of the document according to the pattern of the classification item.   The information retrieval apparatus according to claim 2, wherein the recommendation score calculation unit calculates importance scores of a duplication source document and a duplication destination document according to a pattern of the classification item. .   The recommendation score calculating means includes a document importance calculating means for calculating a score indicating the importance of registered documents, and a document relevance calculating means for calculating a score indicating the degree of association between registered documents. The information search device according to claim 1, wherein: A profile generation step of generating profile information used for recommendation from document operation history information in which operation history for each operation type of a user performed on a plurality of registered documents is recorded;
Analyzing the profile information generated in the profile generation step, and calculating a recommendation score for each of the plurality of documents based on the analysis result; and
A recommendation result display step of displaying a recommendation result document calculated in the recommendation score calculation step on a display device. The profile information has a copy classification table for storing a copy source document and a copy destination document,
In the profile generation step, a document that has been a copy source in the past is extracted from the document operation history information and stored in the copy source document, and the purpose of creating the document from a history indicating that the document has been copied is 12. The information retrieval method according to claim 11, wherein the information is classified into predetermined classification items and stored in the duplication destination document.   13. The information search method according to claim 12, wherein the classification items are classified by estimating what kind of intention a user has created a document from a log record of a copy operation. The profile information has a document relevance table in which all registered documents are stored and relevance between all documents is given,
13. The profile generation step includes updating the relevance information stored in the document relevance table every time a document is used at the same time. Or the information search method of 13.   The profile generation step determines whether or not to take over a document related to a copy source document as a document related to a copy destination document according to a result of classifying the document. Information retrieval method. The profile information has a document importance table storing importance information of all registered documents,
13. The profile generation step assumes that a document viewed or printed by a user is a document having high importance, and updates importance information of the document every time the user browses or prints. The information search method of any one of -14.   The classification items include a pattern for duplicating and updating an old version document, a pattern for duplicating an old version and storing it as a backup, a pattern for duplicating and updating a template, and a pattern for duplicating and dividing a document. The information search method according to any one of claims 12 to 16, wherein the information search method is characterized in that:   The information retrieval method according to claim 12, wherein the recommendation score calculation step calculates an importance score of the document according to the pattern of the classification item.   The information search method according to any one of claims 12 to 18, wherein the recommendation score calculation step calculates importance scores of a duplication source document and a duplication destination document according to a pattern of the classification item. .   The recommendation score calculation step includes a document importance level calculation step for calculating a score indicating the importance level of the registered documents, and a document relevance level calculation step for calculating a score indicating the relevance level between the registered documents. The information search method according to claim 11, wherein: A profile generation step of generating profile information used for recommendation from document operation history information in which operation history for each operation type of a user performed on a plurality of registered documents is recorded;
Analyzing the profile information generated in the profile generation step, and calculating a recommendation score for each of the plurality of documents based on the analysis result; and
A program for causing a computer to execute a recommendation result display step of displaying a recommendation result document calculated in the recommendation score calculation step on a display device.   A computer-readable storage medium storing the program according to claim 21.

Images