Information processing apparatus, information processing system, information processing method, and program

Abstract

To generate a relation diagram between a document file to which a user wants to pay attention, for example, a selected file and another document file by adding a retrieval query input by the user.SOLUTION: The information processing apparatus includes a divided data generation unit configured to generate a plurality of divided data by dividing contents of a plurality of document files, a search unit configured to search the plurality of document files based on a search query input by a user, and a relation diagram data generation unit configured to generate relation diagram data indicating a relation between a selected document file and other document files based on a part of divided data having a relatively high relation to the search query among a plurality of divided data related to the selected document file selected by the user from the searched document files.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present invention relates to an information processing device, an information processing system, an information processing method, and a program. [Background technology]

[0002] Conventionally, there is a system that searches for a group of document files that match a search query from among stored document files.

[0003] Regarding such systems, a technology has been disclosed that generates a relationship diagram in which document files are treated as nodes based on the relationships between the searched document files, and highly related document files are connected by edges (Patent Document 1). Summary of the Invention [Problem to be solved by the invention]

[0004] However, conventional technologies only generate a relationship diagram between searched document files, and therefore cannot generate a relationship diagram between a document file of a user's interest and other document files, taking into account the search query that was the basis for the document search.

[0005] The present invention has been made in consideration of the above points, and aims to generate a relationship diagram between a document file that a user wants to focus on, such as a selected file, and other document files, taking into account a search query entered by the user. [Means for solving the problem]

[0006] In order to solve the above problem, the information processing device has a split data generation unit that generates multiple split data by splitting the contents of multiple document files, a search unit that searches the multiple document files based on a search query entered by a user, and a related diagram data generation unit that generates related diagram data showing the relationship between the selected document file and other document files based on a portion of the multiple split data related to a selected document file selected by the user from the searched document files that has a relatively high relevance to the search query. [Effects of the Invention]

[0007] A relationship diagram between a document file that the user wants to focus on, for example, a selected document file, and other document files can be generated taking into account the search query entered by the user. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 illustrates an example of a configuration of an information processing system according to a first embodiment. [Figure 2] 1 is a diagram illustrating an example of a hardware configuration of an information processing device 10 according to a first embodiment. [Figure 3] FIG. 2 is a diagram illustrating an example of a functional configuration of an information processing system according to a first embodiment. [Figure 4] 10 is a flowchart illustrating an example of a processing procedure executed in response to an update of a document file group. [Figure 5] FIG. 10 is a diagram illustrating an example of the configuration of divided data information. [Figure 6] FIG. 2 is a diagram illustrating an example of the configuration of meta information. [Figure 7] FIG. 10 is a diagram illustrating an example of the configuration of vector information. [Figure 8] FIG. 10 is a diagram illustrating an example of the configuration of cluster information. [Figure 9] FIG. 10 is a diagram illustrating an example of the configuration of word importance information. [Figure 10] FIG. 10 is a sequence diagram illustrating an example of a processing procedure for document file search processing. [Figure 11] FIG. 10 is a diagram showing a display example of a search result screen. [Figure 12] 10 is a flowchart illustrating an example of a processing procedure for generating related chart data. [Figure 13] FIG. 2 is a diagram showing an example of a display of a related diagram in the first embodiment. [Figure 14] FIG. 3 is a diagram for explaining a method for identifying related files in the first embodiment. [Figure 15] 10 is a flowchart illustrating an example of a processing procedure of "divided data-divided data." [Figure 16] FIG. 10 is a diagram for explaining a method for identifying related files in the second embodiment. [Figure 17] 10 is a flowchart illustrating an example of a processing procedure for "cluster-divided data." [Figure 18] FIG. 11 is a diagram showing an example of display of a related diagram in the second embodiment. [Figure 19] FIG. 10 is a diagram showing an example of the configuration of a selected document file for explaining the effect of the second embodiment. [Figure 20] FIG. 13 is a diagram for explaining a method for identifying related files in the third embodiment. [Figure 21] 10 is a flowchart illustrating an example of a processing procedure for "preceding k divided data." [Figure 22] FIG. 13 is a diagram showing an example of display of a related diagram in the third embodiment. [Figure 23] FIG. 13 is a diagram for explaining a method for identifying related files in the fourth embodiment. [Figure 24] 10 is a flowchart illustrating an example of a processing procedure of "Any-cluster." [Figure 25] FIG. 13 is a diagram showing an example of display of a related diagram in the fourth embodiment. [Figure 26] FIG. 13 is a diagram showing an example of display of a related diagram in the fifth embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0009] Hereinafter, embodiments of the present invention will be described with reference to the drawings. Fig. 1 is a diagram showing an example of the configuration of an information processing system in a first embodiment. In Fig. 1, the information processing system includes a file server 20, an information processing device 10, and one or more user terminals 30. The information processing device 10 is connected to the file server 20 via a network N1. The user terminal 30 is connected to the file server 20 via a network N2, and is connected to the information processing device 10 via a network N3.

[0010] The user terminal 30 is a terminal that uploads document files to the file server 20 and requests the information processing device 10 to search for files in response to instructions from a user. For example, a PC (Personal Computer), a tablet terminal, a smartphone, or the like may be used as the user terminal 30.

[0011] A document file is a unit of data management, and the type of data stored in a document file is not limited to a specific one. For example, the data may be a document, an image, or audio. In this embodiment, an example will be described in which data is stored in a document file and distributed.

[0012] The file server 20 is one or more computers that store document files uploaded from the user terminal 30.

[0013] The information processing device 10 is one or more computers that store information about document files stored in the file server 20 and, in response to a file search request from a user terminal 30, searches for a document file corresponding to the search request based on the information. The information processing device 10 also generates relational diagram data showing a diagram (hereinafter referred to as "relational diagram") showing the relationship between a document file selected by a user from the document files included in the search results and other document files, and transmits the relational diagram data to the user terminal 30.

[0014] The file server 20 and the information processing device 10 may be realized using the same computer. In this case, the network N1 corresponds to a signal line such as a bus within the computer that constitutes the file server 20 and the information processing device 10. Alternatively, each user terminal 30 may also function as the information processing device 10. In this case, the network N3 corresponds to a signal line such as a bus within the user terminal 30.

[0015] In the embodiments of the present invention, the scene (situation) in which the information processing system is used is not limited to a specific form, but may be used within a company, for example. That is, each employee of a company (including not only companies but also government agencies, various organizations, unions, etc., and not only full-time employees but also temporary workers, part-time workers, and casual workers) may be a user (in the present embodiment, each employee of a company is described as a user, but this is not limited to this, and the information collection system can also be used by general users). Or it may be an organization in a specific group or region, i.e., the organization in the embodiments of the present invention may be a company in a group of companies or organizations. Note that the following embodiments of the present invention will be described using a scene in a certain company (hereinafter referred to as "Company X") as an example.

[0016] In this case, the file server 20 is a group of computers that store files within the company X. For example, the file server 20 stores various document files created within the company X. In this case, the network N2 corresponds to, for example, a WAN (Wide Area Network) or a LAN (Local Area Network) within the company X.

[0017] The information processing device 10 may be installed within the company X, or may be installed outside the company X (in a cloud environment (e.g., a data center) connected to a network within the company X via the Internet). When the information processing device 10 is installed within the company X, the network N1 and the network N3 correspond to, for example, a wide area network (WAN) or a local area network (LAN) within the company X. When the information processing device 10 is installed within the company X, the network N1 and the network N3 correspond to, for example, the Internet.

[0018] Fig. 2 is a diagram showing an example of the hardware configuration of information processing device 10 in the first embodiment. As shown in Fig. 2, information processing device 10 is constructed by a computer and includes CPU 101, ROM 102, RAM 103, HD 104, HDD (Hard Disk Drive) controller 105, display 106, external device connection I / F (Interface) 108, network I / F 109, data bus 110, keyboard 111, pointing device 112, DVD-RW (Digital Versatile Disk Rewritable) drive 114, and media I / F 116. However, when information processing device 10 is operated via a network, information processing device 10 does not need to include display 106, keyboard 111, pointing device 112, etc.

[0019] The CPU 101 controls the overall operation of the information processing device 10. The ROM 102 stores programs, such as an IPL, used to drive the CPU 101. The RAM 103 is used as a work area for the CPU 101. The HD 104 stores various data, such as programs. The HDD controller 105 controls the reading and writing of various data from and to the HD 104 under the control of the CPU 101. The display 106 displays various information, such as a cursor, menus, windows, characters, or images. The external device connection I / F 108 is an interface for connecting various external devices. In this case, the external devices are, for example, USB (Universal Serial Bus) memories or printers. The network I / F 109 is an interface for data communication using a communication network. The data bus 110 is an address bus, a data bus, or the like, for electrically connecting the components, such as the CPU 101, shown in FIG. 2.

[0020] The keyboard 111 is a type of input means having multiple keys for inputting characters, numbers, various instructions, etc. The pointing device 112 is a type of input means for selecting and executing various instructions, selecting a processing target, moving a cursor, etc. The DVD-RW drive 114 controls reading and writing of various data from a DVD-RW 113, which is an example of a removable recording medium. Note that this is not limited to a DVD-RW, and may be a DVD-R, etc. The media I / F 116 controls reading and writing (storing) of data from a recording medium 115, such as a flash memory.

[0021] The file server 20 and the user terminal 30 may also have the same hardware configuration as that shown in FIG.

[0022] Fig. 3 is a diagram showing an example of the functional configuration of the information processing system according to the first embodiment. In Fig. 3, the file server 20 has a notification unit 21. The notification unit 21 is realized by a process in which one or more programs installed in the file server 20 cause the CPU of the file server 20 to execute the process. The file server 20 also uses a file storage unit 22. The file storage unit 22 can be realized by using, for example, an auxiliary storage device of the file server 20, or a storage device connectable to the file server 20 via a network.

[0023] The file storage unit 22 stores document files uploaded from the user terminal 30. The document files stored in the file storage unit 22 are search targets.

[0024] In response to an update of a document file group stored in the file storage unit 22, the notification unit 21 notifies the information processing device 10 that the document file group has been updated. An update of a document file group refers to, for example, uploading a new document file to the file storage unit 22 (addition of a new document file) or updating an existing document file stored in the file storage unit 22 (changing or adding the content of an existing document file). Therefore, a document file stored in the file storage unit 22 is subject to an update notification at least once when a new document file is uploaded. In the case of uploading a new document file, the document file to be uploaded is the document file that caused the document file group to be updated, and in the case of updating an existing document file, the updated document file is the document file that caused the document file group to be updated. A document file that caused the document file group to be updated is a document file that has been added or updated to the document file group. Alternatively, it can be said to be a document file that has been added to or updated in the document file group.

[0025] The information processing device 10 includes an update notification receiving unit 121, a divided data generating unit 122, a meta information generating unit 123, a vector generating unit 124, a clustering unit 125, a word importance information updating unit 126, a receiving unit 131, a searching unit 132, a related diagram data generating unit 133, and an output unit 134. Each of these units is realized by a process executed by the CPU 101 of one or more programs installed in the information processing device 10. The information processing device 10 also uses an information storage unit 141. The information storage unit 141 can be realized using, for example, the HD 104 or a storage device connectable to the information processing device 10 via a network.

[0026] The update notification receiving unit 121 receives a notification (update notification) that a group of document files has been updated from the file server 20. The update notification includes the document file that caused the update (hereinafter referred to as the "updated file").

[0027] The divided data generation unit 122 generates multiple divided data by dividing the contents of the update file into predetermined units, and records information about each divided data in the information storage unit 141. Because each divided data is a part of the update file, the entire update file can be reproduced by linking the divided data. The divided data may be separated, for example, by a predetermined number of sentences (for example, every three sentences). However, the divided data may also be separated by other units such as the number of characters, the number of words, or a paragraph. Because multiple document files stored in the file storage unit 22 become update files at least once, the divided data generation unit 122 generates divided data for each of all document files stored in the file storage unit 22.

[0028] The meta information generating unit 123 generates meta information for the updated file. The meta information is also called bibliographic information. The meta information generating unit 123 records the generated meta information in the information storage unit 141.

[0029] The vector generation unit 124 uses natural language processing such as BERT to generate a vector (hereinafter referred to as a "semantic vector") indicating the semantic features of each divided data piece of the updated file. The semantic vector may be a vector also known as an embedded or distributed representation. The semantic vector for each divided data piece is referred to as a "divided data vector." The vector generation unit 124 records vector information indicating each divided data vector of the updated file in the information storage unit 141.

[0030] In other words, the semantic vector may be generated by a model obtained through the learning effects of machine learning. Here, machine learning is a technology that allows a computer to acquire human-like learning capabilities, and refers to a technology in which a computer autonomously generates algorithms necessary for judgments such as data classification from training data that is previously acquired, and applies these to new data to make predictions. The learning method for machine learning may be any of supervised learning, unsupervised learning, semi-supervised learning, reinforcement learning, and deep learning, or may be a combination of these learning methods; the learning method for machine learning is not important.

[0031] The clustering unit 125 clusters the plurality of divided data pieces generated for the updated file, thereby classifying (dividing) the plurality of divided data pieces into a plurality of clusters. The clustering unit 125 records information about the generated clusters in the information storage unit 141.

[0032] Every time the update notification receiving unit 121 receives an update notification, the word importance information updating unit 126 updates the word importance information stored in the information storage unit 141. The word importance information is information indicating the importance of each word contained in all document files stored in the file storage unit 22 in each divided data of each document file, and is information for a set of multiple document files stored in the file storage unit 22, rather than for each document file.

[0033] The receiving unit 131 receives (accepts) from the user terminal 30 a document file search request including a search query input by a user of the user terminal 30. The search query is a character string that expresses information to be searched for in a natural language.

[0034] In response to a document file search request from the user terminal 30, the search unit 132 searches for multiple document files (similar to the search query) based on the search query using information stored in the information storage unit 141 regarding multiple document files stored in the file storage unit 22.

[0035] The relationship diagram data generation unit 133 generates relationship diagram data showing a relationship diagram, which is a graphic showing the relationship between the selected document file and other document files (other than the selected document file), based on some of the multiple split data related to the document file (hereinafter referred to as the "selected document file") selected by the user from the document files searched by the search unit 132, that have a relatively high relevance to the search query.

[0036] More specifically, the related diagram data generation unit 133 calculates the similarity between a portion of the split data, which has a relatively high relevance to the search query, among the plurality of split data for a selected document file selected by the user from among the document files searched by the search unit 132, and the split data obtained by dividing the contents of other document files other than the selected document file. Strictly speaking, the related diagram data generation unit 133 calculates the similarity between the split data vector for the portion of the split data and each of the split data vectors for the plurality of split data obtained by dividing the contents of the other document files. For example, cosine similarity may be used as the similarity. In the first embodiment, the portion of the split data is one piece of split data. That is, in the first embodiment, the related diagram data generation unit 133 calculates the similarity between one piece of split data, which is a portion of the split data relatively highly relevance to the search query, among the plurality of split data for a selected document file selected by the user from among the document files searched by the search unit 132, and the split data obtained by dividing the contents of other document files other than the selected document file.

[0037] The relationship diagram data generating unit 133 generates relationship diagram data showing the relationship between the selected document file and other document files related to divided data having a relatively high calculated similarity. The relationship diagram, also called a correlation diagram, is a graphic diagram in the form of a graph in which document files are nodes and the nodes of document files having a relatively high degree of relationship are connected by edges.

[0038] The output unit 134 outputs output information including the search results by the search unit 132 in response to a document file search request, and outputs output information including a related diagram based on the related diagram data generated by the related diagram data generation unit 133. In this embodiment, output means transmitting the related diagram data to the user terminal 30 and displaying it on the user terminal 30.

[0039] The following describes the processing procedure executed in the information processing system: Fig. 4 is a flowchart illustrating an example of the processing procedure executed in response to an update of a document file group.

[0040] The processing procedure in Fig. 4 is initiated when the user terminal 30 adds (newly registers) a document file to the file storage unit 22 of the file server 20 or updates any of the document files stored in the file storage unit 22 in response to a user operation, and the notification unit 21 of the file server 20 sends an update notification to the information processing device 10, indicating that the document files stored in the file storage unit 22 have been updated. The update notification includes the added or updated document file (updated file) that caused the update of the document files, and identification information of the updated file (hereinafter referred to as a "document ID"). The processing procedure in Fig. 4 may be executed after the processing procedure in Fig. 10, which will be described later.

[0041] In step S101, the update notification receiving unit 121 receives the update notification (S101).

[0042] Next, the split data generation unit 122 splits the update file included in the update notification into multiple split data in predetermined units, and records the split data information in which each split data is associated with the document ID included in the update notification in the information storage unit 141 (S102).

[0043] Fig. 5 is a diagram showing an example of the configuration of split data information. As shown in Fig. 5, split data information for one document file includes a document ID for the document file, and a split ID and split data for each split data. The split ID is identification information for the split data. The split data is text data that indicates sentences belonging to the split data among the sentences in the document file.

[0044] If the cause of updating the document file group is the addition of a new document file, the divided data generation unit 122 adds the new divided data information to the information storage unit 141 in association with the document ID of the updated file. If the cause of updating the document file group is the update of an existing document file, the divided data generation unit 122 updates the divided data information in the information storage unit 141 that corresponds to the document ID of the updated file with the newly generated divided data information.

[0045] Next, the meta information generating unit 123 extracts meta information from the update notification and records the meta information in the information storage unit 141 (S103).

[0046] Fig. 6 is a diagram showing an example of the configuration of meta information. As shown in Fig. 6, the meta information for one document file includes the document ID of the document file, the file name (in the file storage unit 22) of the document file, the creation date of the document file, the creator of the document file, the last update date of the document file, and the person who last updated the document file, etc. The file name, the creation date of the document file, the creator of the document file, the last update date of the document file, and the person who last updated the document file, etc. may be extracted from the updated file (the attribute information assigned to the updated file).

[0047] If the cause of updating the document file group is the addition of a new document file, the meta information generation unit 123 adds the new meta information to the information storage unit 141 in association with the document ID of the updated file. If the cause of updating the document file group is the update of an existing document file, the meta information generation unit 123 updates the meta information corresponding to the document ID of the updated file in the information storage unit 141 with the newly generated meta information.

[0048] Next, the vector generation unit 124 generates a divided data vector for each divided data (FIG. 5) corresponding to the updated file recorded in the information storage unit 141 in step S102, and records vector information in which each divided data vector is associated with the division ID and the document ID included in the update notification in the information storage unit 141 (S104). Each divided data (FIG. 5) corresponding to the updated file can be identified based on the document ID included in the update notification.

[0049] Fig. 7 is a diagram showing an example of the configuration of vector information. As shown in Fig. 7, the vector information for one document file includes the document ID of the document file, and a division ID and a divided data vector for each divided data belonging to the document file.

[0050] If the cause of the update of the document file group is the addition of a new document file, the vector generation unit 124 adds a new divided data vector group to the information storage unit 141 in association with the document ID of the updated file. If the cause of the update of the document file group is the update of an existing document file, the vector generation unit 124 updates the divided data vector group corresponding to the document ID of the updated file in the information storage unit 141 with the newly generated divided data vector group.

[0051] Next, the clustering unit 125 classifies the multiple split data vectors generated for the updated file in step S104 into multiple clusters by clustering them using the k-means method or the like, and records cluster information in the information storage unit 141 in which the split ID, the document ID included in the update notification, and identification information of the cluster to which the split data related to the split ID belong (hereinafter referred to as the "cluster ID") (S105).

[0052] Fig. 8 is a diagram showing an example of the configuration of cluster information. As shown in Fig. 8, the cluster information for one document file includes the document ID of the document file, the division ID of each divided data item belonging to the document file, and the cluster ID of the cluster to which the divided data item belongs. Divided data items corresponding to the same cluster ID belong to the same cluster. Note that the number of clusters for one document file is generally smaller than the number of divided data items for the document file.

[0053] If the cause of the update of the document file group is the addition of a new document file, the clustering unit 125 adds new cluster information to the information storage unit 141 in association with the document ID of the updated file. If the cause of the update of the document file group is the update of an existing document file, the clustering unit 125 updates the cluster information corresponding to the document ID of the updated file in the information storage unit 141 with the newly generated cluster information.

[0054] Next, in response to the update instruction, the word importance information update unit 126 updates the word importance information stored in the information storage unit 141 (S106). Specifically, the word importance information update unit 126 calculates the importance of each divided data of the updated file for a set of words (a set of words excluding duplicates) included in all divided data of all document files, including the updated file, and adds the calculation results to the word importance information. Importance refers to an index that indicates the relative importance of each word in a set of words. For example, the TF-IDF value may be used as the importance, or a value of another scale may be used as the importance.

[0055] Fig. 9 is a diagram showing an example of the configuration of word importance information. In Fig. 9, the words arranged in the column direction are a set of words included in all the segmented data of all the document files (a set of words excluding duplicates). Each segmented data of each document file is assigned in the row direction. The value of a certain row and a certain column is the importance of the word related to that column in the segmented data related to that row.

[0056] As a result of the above, the information necessary for searching for an updated file is stored in the information storage unit 141.

[0057] Next, a document file search process will be described with reference to a flowchart of FIG.

[0058] In step S201, the receiving unit 131 receives (accepts) a document file search request from the user terminal 30. The document file search request includes conditions (search conditions) related to the document file search. The search conditions include, for example, a search query, which is a character string expressing the information to be searched for in a natural language.

[0059] A search query is, for example, a set of one or more words. A search query may be a list of one or more words or may have the form of one or more sentences.

[0060] Next, the search unit 132 refers to various pieces of information stored in the information storage unit 141 and executes a process for searching for document files based on the search query.

[0061] First, the search unit 132 generates a semantic vector of the search query (hereinafter referred to as a "search query vector") using the same method as the method for generating the divided data vectors (S202).

[0062] Next, for each document file registered in the information storage unit 141, the search unit 132 calculates the similarity between each divided data vector of the document file and the search query vector, and sets the maximum similarity as the similarity of the document file to the search query (the similarity of the search query to the document file) (S203). As the similarity between the search query vector and the divided data vector, for example, cosine similarity may be used. Note that the search unit 132 stores, for each document file, in the RAM 103, for example, the division ID of the divided data vector with the maximum similarity among the divided data vectors of the document file (hereinafter referred to as the "maximum similarity division ID").

[0063] Next, the search unit 132 identifies the document files with the top N similarities (or the document files with similarities equal to or greater than a threshold) as document files that match the search query (S204).

[0064] Next, the search unit 132 generates search results by acquiring meta-information (FIG. 6) of each identified document file from the information storage unit 141 (S205).

[0065] Next, the output unit 134 generates screen data for a search result screen for displaying the search results, and outputs (transmits) output information including the screen data to the user terminal 30 (S206). The user terminal 30 displays the search result screen based on the screen data.

[0066] 11 is a diagram showing a display example of the search result screen 520. As shown in FIG.

[0067] Search condition display area 521 is an area that displays search conditions, and includes a search query display area 5211. Search query display area 5211 is an area that displays a search query. Note that search query display area 5211 may be operable. In this case, a part or all of the search query may be changed via search query display area 5211, and execute button 5212 may be pressed, thereby re-executing step S201 and subsequent steps in FIG. 10 .

[0068] The search result display area 522 is an area where the creator, updater, file name, label list, and details button 523 are displayed for each document file included in the search results. The updater is, for example, the last updater in the meta information (Fig. 6). The label list is a list of labels extracted from document files. The labels of a certain document file are, for example, the top M words in the word importance information (Fig. 9) with importance values ​​in any of the divided data of the document file. The details button 523 is a button for receiving an instruction to display a relationship diagram related to the document file corresponding to the details button 523.

[0069] By referring to the search result screen 520, the user can check a list of document files found according to the search conditions.

[0070] Next, a description will be given of the process of generating related diagram data, which is executed when an instruction to display a related diagram related to any document file is input via the search result screen 520 in Fig. 11. Fig. 12 is a flowchart for explaining an example of the processing procedure for generating related diagram data.

[0071] When the details button 523 for any document file is pressed (when any document file is selected) on the search result screen 520, the user terminal 30 transmits a request for generating a relationship diagram to the information processing device 10. The request for generation includes the document ID (hereinafter referred to as the "selected document ID") of the document file (selected document file) for which the details button 523 was pressed (i.e., selected).

[0072] In step S301, the receiving unit 131 receives the generation request.

[0073] Next, the related diagram data generating unit 133 executes a process of identifying related files (S302). A related file is a document file that has a relatively high (strong) relationship with the selected document file among all document files other than the selected document file (hereinafter referred to as "other document files") stored in the file storage unit 22. The degree of relationship with the selected document file is calculated using a vector indicating the characteristics of the selected document file (hereinafter referred to as "selected feature vector").

[0074] Next, the related diagram data generating unit 133 generates related diagram data based on the selected document file and the identified related file (S303).

[0075] Next, the output unit 134 generates screen data including the related diagram data, and outputs (transmits) output information including the screen data to the user terminal 30 (S206). The user terminal 30 displays a screen based on the screen data.

[0076] FIG. 13 is a diagram showing a display example of a relationship diagram in the first embodiment. As shown in FIG. 13, the relationship diagram r1 is a graphic in the form of a graph in which files and labels are nodes and relationships between document files (relationships via labels) are edges. In FIG. 13, node nT is a node corresponding to the selected document file (hereinafter referred to as a "selected document file node"). Nodes nR1-1 to 2, node nR2-1, and node nR3-1 to 2 are nodes corresponding to related files (hereinafter referred to as "related file nodes"). Nodes nL1 to 3 are nodes corresponding to labels (hereinafter referred to as "label nodes"). When the label nodes nL1 to 3 are not distinguished from one another, they are referred to as "label node nL." When the related file nodes nR1-1 to 2, nR2-1, and nR3-1 to 2 are not distinguished from one another, they are referred to as "related file node nR."

[0077] The "label" of the label-mediated association is a word that indicates in what sense or perspective the associated file is associated with the selected document file. The method of extracting the label will be described later.

[0078] Next, the process of identifying related files in step S302 will be described. The procedure for identifying related files differs depending on the embodiment.

[0079] FIG. 14 is a diagram for explaining a method for identifying related files in the first embodiment.

[0080] FIG. 14 shows that the similarity is calculated for each pair of one piece of divided data (hereinafter referred to as "target divided data") in the selected document file and all the pieces of divided data in the other document files. That is, in the first embodiment, the related diagram data generation unit 133 calculates the similarity between one piece of divided data (hereinafter referred to as "target divided data") as a part of divided data that has a relatively high relevance to the search query and multiple pieces of divided data obtained by dividing the contents of the other document files. Specifically, the related diagram data generation unit 133 sets the divided data vector of the target divided data (hereinafter referred to as "target divided data vector") as a selected feature vector, and calculates the similarity for each pair of the selected feature vector and each divided data vector of the other document files. The related diagram data generation unit 133 sets the other document files related to the divided data with a relatively high similarity as related files, and generates related diagram data showing the relationship between each related file and the selected document file. Here, the target split data refers to the split data among the multiple split data related to the selected document file that has the highest similarity between its split data vector and the search query vector. Therefore, the target split data can be identified based on the split ID with the highest similarity to the selected document file. Note that the method of identifying related files by calculating the similarity between split data, as shown in Figure 14, is called "split data-split data."

[0081] FIG. 15 is a flowchart illustrating an example of a procedure for processing "divided data-divided data."

[0082] In step S410, the relational diagram data generating unit 133 acquires the target divided data vector as a selected feature vector.

[0083] Next, the relationship diagram data generating unit 133 calculates the similarity (for example, cosine similarity) between the selected feature vector and the divided data vectors of all the divided data of each of the other document files (S420).

[0084] Next, the relationship diagram data generation unit 133 identifies, for each other document file, the divided data vector with the highest similarity among the divided data of the other document file (S431). This process is for establishing a one-to-one correspondence between the divided data vector and the other document file. That is, as a result of this process, there will be only one divided data vector (divided data) related to each other document file.

[0085] Next, the related diagram data generation unit 133 identifies the top M divided data vectors with the highest similarity from the divided data vectors identified for each of the other document files (S432). As a result, the other document files including divided data related to the identified divided data vectors are identified as related files for the selected document file.

[0086] Next, the relationship diagram data generation unit 133 classifies the top M divided data vectors into multiple clusters by clustering them using the k-means method or the like (S433). Since the top M divided data vectors and other document files (related files) have a one-to-one correspondence, each cluster corresponds to a cluster that classifies M related files. Hereinafter, the clusters generated in step S433 will be referred to as "related clusters."

[0087] Next, the relational diagram data generating unit 133 executes a loop process including steps S441 to S443 for each relational cluster. The relational cluster that is the processing target in this loop process is referred to as a "target relational cluster."

[0088] In step S441, the relationship diagram data generation unit 133 identifies one characteristic word in the target related cluster as a label. For example, the relationship diagram data generation unit 133 identifies the most important word among the words contained in each related file (hereinafter referred to as "other related files belonging to the target related cluster") related to the divided data vector belonging to the target related cluster by referring to the word importance information (FIG. 9). The label identified here becomes one label node in the relationship diagram (FIG. 13). Therefore, the related files belonging to the target related cluster are connected to the selected document file via the label.

[0089] Next, the relational diagram data generating unit 133 calculates a feature vector of the target-related cluster (S442). The feature vector may be, for example, a vector as the center of gravity of the divided data vectors belonging to the target-related cluster, or a vector using the maximum value for each dimension of the divided data vectors belonging to the target-related cluster.

[0090] Next, the relational diagram data generation unit 133 calculates the similarity (e.g., cosine similarity) between the selected feature vector and the feature vector of the target related cluster (S443). In the first embodiment, the selected feature vector is the target divided data vector. Therefore, the similarity has the meaning of the similarity between the target divided data and the target related cluster.

[0091] When steps S441 to S443 have been executed for all related clusters, the processing procedure in FIG. 15 ends.

[0092] In step S303 of FIG. 12, the relationship diagram data generating unit 133 generates relationship diagram data based on the related files identified in FIG. 15 and the labels and similarities identified for each related cluster.

[0093] Referring again to FIG. 13 , for example, the relationship diagram data generation unit 133 places the selected document file node nT in the relationship diagram r1. Next, the relationship diagram data generation unit 133 places a label node nL corresponding to each related cluster and connects the selected document file node nT to each label node nL with an edge. At this time, the relationship diagram data generation unit 133 determines, for each label node nL, the color of the label node nL and the edge, the length of the edge (the distance from node nT), the thickness of the edge, the line type of the edge, etc., based on the similarity between (the feature vector of) the related cluster corresponding to the label node nL and (the selected feature vector of) the selected document file. The relationship diagram data generation unit 133 may darken the color of the label node nL and the edge as the similarity increases, shorten the edge length as the similarity increases, and thicken the edge as the similarity increases. If the similarity is higher than a threshold, the edge may be drawn with a solid line, or if not, the edge may be drawn with a dashed line. The relationship diagram data generation unit 133 also arranges, for each label node nL, a related file node nR corresponding to a related file belonging to the related cluster corresponding to the label node nL, and connects the related file node nR to the label node nL with an edge. The relationship diagram data generation unit 133 sets the color, length, thickness, and line type of the edge between the label node nL and the related file node nR to be the same as, for example, the edge between the label node nL and the selected document file node. The relationship diagram data generation unit 133 also sets the color of each related file node nR to be the same as, for example, the color of the label node nL to which the related file node nR is connected. By generating the relationship diagram in this manner, it is possible to visually represent the degree of relevance between the selected document file and each related cluster.

[0094] Furthermore, the related diagram data generation unit 133 assigns a symbol m1, which is a pair of an upside-down "" and a "", to the selected document file node nT and all related file nodes nR as information indicating that the related file identification method is "split data-split data". The symbol m1 may be represented by an icon. The symbol m1 indicates that for the document file corresponding to the node to which the symbol is assigned, the relevance (similarity) with the selected document file or other document files has been evaluated using one split data as a comparison unit. In FIG. 13, the symbol m1 is assigned to the selected document file node nT and all related file nodes nR, so the user can know that the related files have been identified by "split data-split data".

[0095] Furthermore, the related diagram data generator 133 may generate related diagram data so that the contents of the divided data corresponding to the node that is the target of operation are displayed by a predetermined operation (e.g., mouse-over, click, etc.) on the selected document file node nT and each related file node nR. The example in Fig. 13 shows an example in which text t1 indicating the contents of the target divided data is displayed when the selected document file node nT is operated.

[0096] Furthermore, the relationship diagram data generation unit 133 may add a search query q1 to the relationship diagram r1. By doing so, it is possible to confirm the search query that is the basis of the relationship diagram r1, and to know what search query the selected document file and each related file are similar to (highly related to) each other.

[0097] As described above, according to the first embodiment, relationship diagram data is generated based on target segment data that is most relevant to the search query in a selected document file selected by a user from multiple document files searched based on the search query. That is, other document files containing segment data that are relatively highly similar to the target segment data are arranged in the relationship diagram as related files. As a result, it is possible to grasp the relationship between document files based on the degree of relevance to the search query. That is, a relationship diagram between a document file that the user wants to focus on, such as a selected document file, and other document files can be generated taking into account the search query entered by the user. Creating such a relationship diagram taking into account the search query makes it easier for the user to find the document file they are looking for.

[0098] Furthermore, in the first embodiment, for a selected document file, the split data that is the target for calculating the similarity with all split data of other document files is one split data (target split data), so the amount of calculation can be reduced compared to when calculating the similarity for all split data of the selected document file.

[0099] Next, a second embodiment will be described. In the second embodiment, differences from the first embodiment will be described. Therefore, unless otherwise specifically mentioned, the second embodiment may be the same as the first embodiment. In the second embodiment, the method of identifying related files differs from the first embodiment.

[0100] FIG. 16 is a diagram for explaining a method for identifying related files in the second embodiment.

[0101] 16 shows that the similarity is calculated for each pair of a cluster to which a target divided data piece of the selected document file belongs and all divided data pieces of the other document files. That is, in the second embodiment, the related diagram data generation unit 133 calculates the similarity between a cluster (hereinafter referred to as a "target cluster") to which some divided data pieces (target divided data pieces) having a relatively high relevance to the search query belong, among multiple clusters obtained by clustering multiple divided data pieces related to the selected document file, and multiple divided data pieces related to the other document files. Specifically, the related diagram data generation unit 133 sets the feature vector of the target cluster as a selected feature vector, and calculates the similarity for each pair of the selected feature vector and each divided data vector of the other document files. The related diagram data generation unit 133 sets the other document files related to the divided data pieces having a relatively high similarity as related files, and generates the related diagram data showing the relationship between each related file and the selected document file. As shown in FIG. 16, the method of identifying related files by calculating the similarity between the target cluster and all divided data of other document files is called "cluster-divided data".

[0102] In the second embodiment, the process of identifying related files in step S302 is realized by, for example, the processing procedure shown in FIG.

[0103] Fig. 17 is a flowchart for explaining an example of the processing procedure for "cluster-divided data". In Fig. 17, steps that are the same as or correspond to those in Fig. 15 are given the same reference numerals, and their explanations will be omitted as appropriate. In Fig. 17, step S410 in Fig. 15 is replaced by steps S411 and S412.

[0104] In step S411, the related diagram data generating unit 133 acquires each of the divided data that belong to the same cluster (target cluster) as the target divided data from among the clusters indicated by the cluster information (FIG. 8) of the selected document file. Note that clustering of the selected document file may be performed at this timing.

[0105] In step S412, the relationship diagram data generation unit 133 calculates a feature vector of the target cluster based on the divided data vectors of each divided data belonging to the target cluster. This feature vector is set as a selected feature vector. The feature vector of the target cluster may be, for example, a vector representing the center of gravity of the divided data vectors belonging to the target cluster, or a vector using the maximum value for each dimension of the divided data vectors belonging to the target cluster.

[0106] Thereafter, the feature vector of the target cluster is set as the selected feature vector (the feature vector of the selected document file) and the same processing procedure as in Fig. 15 is executed. As a result, in step S303 in Fig. 12, for example, relationship diagram data showing the following relationship diagram is generated.

[0107] Fig. 18 is a diagram showing an example of a display of a related diagram in the second embodiment. In Fig. 18, parts that are the same as or correspond to those in Fig. 13 are given the same reference numerals, and their explanation will be omitted as appropriate.

[0108] In the second embodiment, when generating the relationship diagram data of the relationship diagram r2, the relationship diagram data generation unit 133 assigns the symbol m2 to the selected document file node nT instead of the symbol m1. The symbol m2 may be represented by an icon. The symbol m2 indicates that the document file corresponding to the node to which the symbol is assigned has been evaluated for its relevance (similarity) with other document files, using the cluster to which one segmented data (target segmented data) belongs as the comparison unit. In FIG. 18, the symbol m2 is assigned to the selected document file node nT, and the symbol m1 is assigned to all related file nodes nR, so the user can know that related files have been identified by the "cluster-segmented data."

[0109] Furthermore, the relationship diagram data generation unit 133 may generate relationship diagram data such that the contents of the divided data corresponding to the selected document file node nT and each related file node nR are displayed in response to a predetermined operation (e.g., mouse-over or click) on the selected document file node nT and each related file node nR. In the second embodiment, the divided data corresponding to the selected document file node nT are multiple divided data belonging to the target cluster. Therefore, the example in FIG. 18 shows an example in which, upon operation of the selected document file node nT, text t2 indicating the contents of the multiple divided data belonging to the target cluster is displayed. In this case, since the text t2 may be long, the relationship diagram data generation unit 1332 may provide a summary button b21 so that a summary of the contents of the text t2 is displayed in response to operation of the summary button b21. The summary can be generated using a known natural language processing technique.

[0110] Note that the configuration of the relational diagram r2 in Fig. 18 is the same as that of the relational diagram r1 in Fig. 13, but in the second embodiment, the related files are identified based on the feature vectors of the target cluster, so the identified related files are not necessarily the same as those in the first embodiment. Therefore, the configuration of the relational diagram r2 is not necessarily the same as that of the relational diagram r1.

[0111] As described above, according to the second embodiment, relationship diagram data is generated based on a target cluster to which target segment data that is most relevant to the search query belongs in a selected document file selected by a user from among multiple document files searched based on the search query. That is, other document files that include segment data that are relatively highly similar to the target cluster are arranged in the relationship diagram as related files. As a result, it is possible to grasp the relationship between document files based on the degree of relevance to the search query.

[0112] Furthermore, in the second embodiment, by identifying related files based on the target cluster, it is possible to identify related files that are similar to the selected document file in a broader sense than in the first embodiment, which may result in identifying more related files.

[0113] For example, suppose a document file such as that shown in FIG. 19 is the selected document file, and divided data 4 ("A's Impressions (Fun)") is the target divided data. Note that "divided data X" indicates divided data with a division ID of X. In the first embodiment, other document files containing divided data relating to fun impressions are identified as related files based on the meaning of such target divided data (divided data vector). On the other hand, in the second embodiment, other document files containing divided data similar to divided data 6 ("B's Impressions (Exciting)") or divided data 8 ("C's Impressions (Lingering Impression)"), which belong to the same cluster as the target divided data, are identified as related files and can be included in the relationship diagram. In this way, the second embodiment is suitable when you want to search for document files more widely.

[0114] Next, a third embodiment will be described. In the third embodiment, differences from the first embodiment will be described. Therefore, unless otherwise specifically mentioned, the third embodiment may be the same as the first embodiment. In the third embodiment, the method of identifying related files differs from the first embodiment.

[0115] FIG. 20 is a diagram for explaining a method for identifying related files in the third embodiment.

[0116] FIG. 20 illustrates that the similarity is calculated for each pair of consecutive split data between a target split data and k consecutive split data before and after the target split data among the split data of the selected document file, and all split data of other document files. The "k before and after" refers to k split data before and k split data after the target split data, and FIG. 20 illustrates an example where k=1. That is, in the third embodiment, the related diagram data generation unit 133 calculates the similarity between consecutive split data (hereinafter referred to as "target consecutive split data") including one split data (target split data) that has a relatively high relevance to the search query and multiple split data obtained by dividing the contents of other document files. Specifically, the related diagram data generation unit 133 selects the feature vector of the target consecutive split data as a selected feature vector and calculates the similarity for each pair of the selected feature vector and each split data vector of the other document files. The related diagram data generation unit 133 generates related diagram data indicating the relevance of other document files to split data with a relatively high similarity to the selected document file. Note that the number of split data before and after the target consecutive split data does not need to be the same. Furthermore, the target continuous divided data may be a plurality of continuous divided data pieces that includes the target divided data piece at the beginning, or may be a plurality of continuous divided data pieces that includes the target divided data piece at the end.

[0117] As shown in FIG. 20, the method of identifying related files by calculating the similarity between the target continuous divided data and all divided data of other document files is called "preceding k divided data".

[0118] In the third embodiment, the process of identifying related files in step S302 is realized by, for example, the processing procedure shown in FIG.

[0119] Fig. 21 is a flowchart for explaining an example of the processing procedure of "preceding k divided data". In Fig. 21, steps that are the same as or correspond to those in Fig. 15 are given the same reference numerals, and their explanation will be omitted as appropriate. In Fig. 21, step S410 in Fig. 15 is replaced by step S413.

[0120] In step S413, the related diagram data generation unit 133 calculates feature vectors of multiple consecutive split data (target consecutive split data) from the split data of the selected document file, which are consecutive between the target split data and k split data before and after the target split data. The feature vectors may be, for example, vectors representing the center of gravity of the multiple split data vectors included in the target consecutive split data, or vectors that adopt the maximum value for each dimension of the multiple split data vectors included in the target consecutive split data.

[0121] Thereafter, the feature vector of the target continuous division is set as the selected feature vector (the feature vector of the selected document file), and the same processing procedure as in Fig. 15 is executed. As a result, in step S303 in Fig. 12, association diagram data showing, for example, the following association diagram is generated.

[0122] Fig. 22 is a diagram showing an example of a display of a related diagram in the third embodiment. In Fig. 22, parts that are the same as or correspond to those in Fig. 13 are given the same reference numerals, and their explanation will be omitted as appropriate.

[0123] In the third embodiment, when generating the relationship diagram data of the relationship diagram r3, the relationship diagram data generation unit 133 assigns the symbol m3 to the selected document file node nT instead of the symbol m1. The symbol m3 may be represented by an icon. The symbol m3 indicates that the document file corresponding to the node to which the symbol is assigned has been evaluated for its relevance (similarity) with other document files using multiple consecutive divided data as a comparison unit. In FIG. 22, the symbol m3 is assigned to the selected document file node nT, and the symbol m1 is assigned to all related file nodes nR, so the user can know that related files have been identified by "the k divided data."

[0124] Furthermore, the relationship diagram data generation unit 133 may generate relationship diagram data such that the contents of the divided data corresponding to the selected document file node nT and each related file node nR are displayed in response to a predetermined operation (e.g., mouseover or click) on the selected document file node nT and each related file node nR. In the third embodiment, the divided data corresponding to the selected document file node nT is the target continuous divided data. Therefore, the example in FIG. 22 illustrates an example in which, upon operation of the selected document file node nT, text t3 indicating the contents of multiple divided data (2k+1 divided data) included in the target continuous divided data is displayed. In this case, since the text t3 may be long, the relationship diagram data generation unit 1333 may provide a summary button b31 so that a summary of the contents of the text t3 is displayed in response to operation of the summary button b31. The summary can be generated using a known natural language processing technique.

[0125] Furthermore, the related diagram data generation unit 133 may display a setting area p1 when the setting button s1 is pressed. In the setting area p1, the value of k in the "k items before and after" can be changed. After changing the value of k, when the redraw button b32 is pressed, the process of FIG. 21 is executed again, and the related diagram data is regenerated.

[0126] As in the second embodiment, the configuration of relational diagram r3 is not necessarily the same as relational diagram r1.

[0127] According to the third embodiment, relationship diagram data is generated based on a plurality of consecutive divided data (target consecutive divided data) including the target divided data that is most relevant to the search query in a selected document file selected by a user from among a plurality of document files searched based on the search query. That is, other document files including divided data that are relatively highly similar to the target consecutive divided data are arranged as related files in the relationship diagram. As a result, it is possible to grasp the relationship between document files based on the degree of relevance to the search query.

[0128] Furthermore, in the third embodiment, by identifying related files based on the target continuous divided data, it is possible to identify related files taking into consideration the context before and after the target divided data.

[0129] Next, a fourth embodiment will be described. In the fourth embodiment, differences from the above-described embodiments will be described. Therefore, unless otherwise specified, the fourth embodiment may be the same as the above-described embodiments.

[0130] In the fourth embodiment, the comparison unit for the selected document file is either the target divided data, the target cluster, or the target consecutive divided data. On the other hand, the comparison unit for the related file is the cluster ( FIG. 8 ) generated by clustering the multiple divided data into which the related file is divided. That is, in the fourth embodiment, the related diagram data generation unit 133 calculates the similarity between the target divided data, the target cluster, or the target consecutive divided data for each cluster obtained by clustering the multiple divided data obtained by dividing the contents of another document file for each other document file. More specifically, for each cluster obtained by clustering the multiple divided data obtained by dividing the contents of another document file for each other document file, the related diagram data generation unit 133 calculates the similarity between the feature vector of the cluster and a selected feature vector based on the target divided data, the target cluster, or the target consecutive divided data. The related diagram data generation unit 133 generates related diagram data indicating the relationship between the selected document file and other document files related to clusters with a relatively high similarity.

[0131] As shown in FIG. 23, a method of specifying related files by calculating the similarity between the target divided data, the target cluster, or the target continuous divided data and all clusters of other document files is called "Any-cluster."

[0132] In the fourth embodiment, the process of identifying related files in step S302 is realized by, for example, the process procedure shown in FIG.

[0133] Fig. 24 is a flowchart for explaining an example of the processing procedure of "Any-cluster". In Fig. 24, steps that are the same as or correspond to those in Fig. 15, Fig. 17, or Fig. 21 are given the same reference numerals, and their explanations will be omitted as appropriate. In Fig. 24, step S410 in Fig. 15 is replaced by step S414.

[0134] In step S414, the related diagram data generation unit 133 acquires a selected feature vector. When the divided data vector of the target divided data is set as the selected feature vector (similar to the first embodiment), step S410 in FIG. 15 is executed in step S414. When the feature vector of the cluster to which the target divided data belongs is set as the selected feature vector (similar to the second embodiment), steps S411 and S412 in FIG. 17 are executed in step S414. When the feature vector of the target continuous divided data is set as the selected feature vector (similar to the third embodiment), step S413 in FIG. 21 is executed in step S414.

[0135] Subsequently, steps S421 to S423 are executed for each other document file instead of step S420 in Fig. 15. The other document file that is the processing target in the loop processing for each other document file (hereinafter referred to as "loop processing 1") is referred to as the "target other file."

[0136] In step S421, the relationship diagram data generating unit 133 acquires the clusters of the target other files from the cluster information (FIG. 8). However, in step S421, the target other files may be clustered.

[0137] Next, the relationship diagram data generation unit 133 executes steps S422 and S423 for each cluster of the target other file. A cluster that is the processing target in the loop processing for each cluster of the target other file (hereinafter referred to as "loop processing 2") is referred to as a "target other cluster."

[0138] In step S422, the relationship diagram data generating unit 133 calculates a feature vector of the target other cluster. The feature vector may be, for example, a vector as the center of gravity of the divided data vectors belonging to the target other cluster, or a vector using the maximum value for each dimension of the divided data vectors belonging to the target other cluster.

[0139] Next, the relationship diagram data generating unit 133 calculates the similarity between the selected feature vector and the feature vector of the target other cluster (S423).

[0140] When the loop process 2 is executed for all clusters of the target other file, the similarity is obtained for each cluster of the target other file.

[0141] When the loop process 1 is executed for all other document files, the similarity is obtained for each cluster of all other document files.

[0142] Subsequently, steps S431a to S433a are executed instead of steps S431 to S433. In step S431a, the relationship diagram data generating unit 133 identifies, for each other document file, the feature vector of the cluster having the highest similarity among the clusters of the other document file. Next, the relationship diagram data generating unit 133 identifies the feature vectors of the clusters with the highest similarity from the feature vectors of the clusters identified for each of the other document files (S432a). As a result, the other document files including the clusters related to the identified feature vectors are identified as related files for the selected document file.

[0143] Next, the relationship diagram data generation unit 133 classifies the top M feature vectors into multiple clusters by clustering them using the k-means method or the like (S433a). Since there is a one-to-one correspondence between the top M feature vectors and other document files (related files), each cluster corresponds to a cluster of related files (related cluster).

[0144] Next, the relational diagram data generating unit 133 executes a loop process including steps S441a to S443a for each relational cluster. The relational cluster that is the processing target in this loop process is referred to as a "target relational cluster."

[0145] In step S441a, the relationship diagram data generation unit 133 identifies one characteristic word in the target related cluster as a label. For example, the relationship diagram data generation unit 133 refers to the word importance information (FIG. 9) to identify the most important word among the words contained in each related file (hereinafter referred to as "other related files belonging to the target related cluster") related to the feature vector of a cluster (cluster of other document files) belonging to the target related cluster. The label identified here becomes one label node in the relationship diagram. Therefore, the related file belonging to the target related cluster is connected to the selected document file via the label.

[0146] Next, the relationship chart data generation unit 133 calculates a feature vector of the target-related cluster (S442a). The feature vector may be, for example, a vector as the center of gravity of the feature vectors of each cluster belonging to the target-related cluster, or a vector using the maximum value for each dimension of the feature vectors of each cluster belonging to the target-related cluster.

[0147] Next, the relational diagram data generating unit 133 calculates the similarity (for example, cosine similarity) between the selected feature vector and the feature vector of the target relational cluster (S443a).

[0148] When steps S441a to S443a have been executed for all related clusters, the processing procedure in FIG. 24 ends.

[0149] Based on the related files thus identified, in step S303 of FIG. 12, for example, related diagram data showing the following related diagram is generated.

[0150] Fig. 25 is a diagram showing an example of a display of a related diagram in the fourth embodiment. In Fig. 25, parts that are the same as or correspond to those in Fig. 22 or Fig. 18 are given the same reference numerals, and their explanation will be omitted as appropriate.

[0151] In the fourth embodiment, when generating the relationship diagram data of the relationship diagram r4, the relationship diagram data generation unit 133 assigns the symbol m2 described in FIG. 18 to each related file node nR instead of the symbol m1. Therefore, the user can know that the related file has been identified by "Any-cluster." Note that the selected document file node nT in the relationship diagram r4 in FIG. 25 is assigned the symbol m3 described in FIG. 22. Therefore, the user can know that the related file has been identified by "k clusters before and after." That is, FIG. 25 shows an example in which the feature vector of the target continuous divided data is used as the selected feature vector.

[0152] Furthermore, the related diagram data generation unit 133 may display a setting area p2 by pressing the setting button s1, allowing the user to select either "prefer accuracy" or "prefer speed." The setting area p2 includes a radio button r41 in addition to an area for changing the value of k in the "prefer k before and after" field. The radio button r41 includes options for "prefer accuracy" and "prefer speed." "Prefer accuracy" is an option corresponding to a method in which the unit of comparison between related files and the selected document file is segment data, as in the first to third embodiments. "Prefer speed" is an option corresponding to a method in which the unit of comparison between related files and the selected document file is cluster data, as described in the fourth embodiment. Because the number of all clusters of all other document files is smaller than the number of all segment data of all other document files, comparison by cluster unit is "prefer speed." When a setting is made in the setting area p2 and the redraw button b32 is pressed, processing according to the setting is executed and the related diagram data is regenerated. Therefore, when the feature vector of the target continuous divided data of the selected document file is the selected feature vector as shown in Figure 25, if "Accuracy Priority" is selected in the radio button r41, a relationship diagram will be generated in the same way as in the third embodiment.

[0153] As described above, according to the fourth embodiment, the amount of calculation can be reduced compared to the above embodiments.

[0154] Next, a fifth embodiment will be described. In the fifth embodiment, differences from the above-described embodiments will be described. Therefore, unless otherwise specified, the fifth embodiment may be the same as the above-described embodiments.

[0155] In the fifth embodiment, in step S303 of FIG. 12, the related chart data generating unit 133 generates related chart data showing a related chart such as that shown in FIG.

[0156] Fig. 26 is a diagram showing an example of a display of a related diagram in the fifth embodiment. In Fig. 26, the same parts as in Fig. 13 are given the same reference numerals, and the description thereof will be omitted.

[0157] In the fifth embodiment, the related diagram data generation unit 133 generates related diagram data so that a setting area p3 is displayed when any one of the menus in the menu list displayed by pressing the setting button s1 is selected. The setting area p3 includes a radio button b51 and a redraw button b52. The radio button b51 is a radio button for selecting the method of identifying the selected document file from among "split data-split data" (first embodiment), "cluster-split data" (second embodiment), and "preceding k pieces-split data" (third embodiment), and includes three options.

[0158] The first option corresponds to "split data-split data" (first embodiment). The second option corresponds to "cluster-split data" (second embodiment). The third option corresponds to "k pieces before and after-split data" (third embodiment). When the user selects one of the options and presses the redraw button b52, the information processing device 10 generates related diagram data based on the method for identifying related files corresponding to the selected option.

[0159] In this way, in the third embodiment, the first to third embodiments can be switched by the user's selection.

[0160] The related diagram data generating unit 133 may also generate related diagram data so that the related diagram data is displayed in the setting area p2 shown in Fig. 25 when any menu is selected from the menu list displayed by pressing the setting button s1. In this way, the fourth embodiment can be further implemented in accordance with the user's selection.

[0161] As described above, according to the fifth embodiment, it is possible to generate an association diagram that matches the user's usage scene.

[0162] The information processing device 10 or the user terminal 30 may be, for example, an image forming device, a PJ (Projector), an IWB (Interactive White Board: an electronic white board with a blackboard function that allows mutual communication), an output device such as digital signage, a HUD (Head Up Display) device, industrial machinery, an imaging device, a sound collection device, medical equipment, a network home appliance, a notebook PC (Personal Computer), a mobile phone, a smartphone, a tablet terminal, a game console, a PDA (Personal Digital Assistant), a digital camera, a wearable PC or a desktop PC, etc.

[0163] Each function in each of the above embodiments can be realized by one or more processing circuits. Here, the term "processing circuit" in this specification includes a processor programmed to execute each function by software, such as a processor implemented by an electronic circuit, as well as devices such as an ASIC (Application Specific Integrated Circuit), a DSP (Digital Signal Processor), an FPGA (Field Programmable Gate Array), and conventional circuit modules designed to execute each of the above-described functions.

[0164] Additionally, the devices described in the above embodiments are merely illustrative of one of multiple computing environments for implementing the embodiments disclosed herein.

[0165] In one embodiment, information processing apparatus 10 includes multiple computing devices, such as a server cluster, configured to communicate with each other over any type of communications link, including a network, shared memory, etc., to perform the processing disclosed herein. Similarly, user terminal 30 may include multiple computing devices configured to communicate with each other.

[0166] Although the embodiments of the present invention have been described in detail above, the present invention is not limited to such specific embodiments, and various modifications and variations are possible within the scope of the gist of the present invention as described in the claims.

[0167] For example, aspects of the present invention are as follows.

[0168] <1> a divided data generation unit that generates a plurality of divided data by dividing the contents of a plurality of document files; a search unit that searches the plurality of document files based on a search query input by a user; a related diagram data generating unit that generates related diagram data showing the relationship between the selected document file and other document files based on a portion of divided data that has a relatively high relevance to the search query among a plurality of divided data related to a selected document file selected by the user from the searched document files; An information processing device comprising:

[0169] <2> the related diagram data generation unit calculates a similarity between one of the divided data as the part of the divided data and each of the plurality of divided data related to the other document file, and generates the related diagram data indicating a relationship between the other document file and the divided data having a relatively high similarity. Characterized by <1> The information processing device described in

[0170] <3> the related diagram data generation unit calculates a similarity between a cluster to which some of the divided data belongs, among a plurality of clusters obtained by clustering a plurality of divided data related to the selected document file, and each of a plurality of divided data related to the other document files, and generates the related diagram data indicating a relationship between the other document files and the selected document file for the divided data having a relatively high similarity; Characterized by <1> The information processing device described in

[0171] <4> the relationship diagram data generation unit calculates a similarity between a plurality of consecutive divided data sets including one divided data set having a relatively high relevance to the search query and each of a plurality of divided data sets related to the other document file, and generates the relationship diagram data indicating the relevance between the other document file and the divided data set having a relatively high similarity, and the selected document file; Characterized by <1> The information processing device described in

[0172] <5> the related diagram data generation unit calculates the similarity for each cluster obtained by clustering the plurality of divided data related to the other document file for each of the other document files, and generates the related diagram data indicating the relationship between the other document file and the selected document file for the other document file related to a cluster having a relatively high similarity; Characterized by <2> ~ <4> 10. The information processing device according to claim 9, wherein:

[0173] <6> a vector generation unit that generates, for each of the divided data, a divided data vector that indicates semantic features of the divided data; and the search unit searches the plurality of document files for document files associated with the divided data vectors that have a relatively high similarity to a search query vector that indicates semantic features of the search query; the part of the split data is split data relating to a split data vector that has a relatively high similarity to the search query vector among the plurality of split data relating to the selected document file; Characterized by <1> ~ <5> 10. The information processing device according to claim 9, wherein:

[0174] <7> the relationship diagram data generation unit calculates a similarity between a divided data vector relating to the part of the divided data and each divided data vector of the plurality of divided data relating to the other document file, classifies the divided data vectors of the plurality of divided data relating to the other document file having a relatively high similarity into a plurality of clusters by clustering, and generates the relationship diagram data showing the relationship between the selected document file and the clusters. Characterized by <1> ~ <6> 10. The information processing device according to claim 9, wherein:

[0175] <8> the related chart data generating unit generates the related chart data so as to include information indicating that the related chart data is based on the part of the divided data. Characterized by <1> ~ <7> 10. The information processing device according to claim 9, wherein:

[0176] <9> The information indicating that the information is based on the partial divided data is an icon. Characterized by <8> The information processing device described in

[0177] <10> the relationship diagram data generation unit generates the relationship diagram data so as to include an icon indicating that the similarity has been calculated using the cluster related to the other document file, the icon being different from an icon indicating that the similarity is calculated based on the part of the divided data. Characterized by <5> The information processing device described in

[0178] <11> the related diagram data generation unit is capable of selecting, when calculating the similarity between the selected document file and the other document file, a method of calculating the similarity for each cluster obtained by clustering a plurality of divided data related to the other document file for each of the other document files, or a method of calculating the similarity using a portion of the divided data of the other document file; Characterized by <5> The information processing device described in

[0179] <12> further comprising an output unit that displays the related chart data generated by the related chart data generation unit on a terminal. Characterized by <1> ~ <11> 10. The information processing device according to claim 9, wherein:

[0180] <13> the output unit causes the terminal to display the related diagram data showing the contents of the part of the divided data. Characterized by <12> The information processing device described in

[0181] <14> a divided data generation unit that generates a plurality of divided data by dividing the contents of a plurality of document files; a search unit that searches the plurality of document files based on a search query input by a user; a related diagram data generating unit that generates related diagram data showing the relationship between the selected document file and other document files based on a portion of divided data that has a relatively high relevance to the search query among a plurality of divided data related to a selected document file selected by the user from the searched document files; an information processing device having a terminal that displays a relational diagram based on the relational diagram data; An information processing system comprising:

[0182] <15> a divided data generation procedure for generating a plurality of divided data by dividing the contents of a plurality of document files; a search step of searching the plurality of document files based on a search query input by a user; a related diagram data generating step of generating related diagram data showing the relationship between the selected document file and other document files based on a portion of divided data that has a relatively high relevance to the search query among a plurality of divided data related to a selected document file selected by the user from the searched document files; An information processing method characterized by being executed by a computer.

[0183] <16> a divided data generation procedure for generating a plurality of divided data by dividing the contents of a plurality of document files; a search step of searching the plurality of document files based on a search query input by a user; a related diagram data generating step of generating related diagram data showing the relationship between the selected document file and other document files based on a portion of divided data that has a relatively high relevance to the search query among a plurality of divided data related to a selected document file selected by the user from the searched document files; A program that causes a computer to execute the following. [Explanation of symbols]

[0184] 10. Information processing equipment 20 File Server 21 Notification Department 22 File storage section 30 User terminals 121 Update notification receiving unit 122 Split data generation unit 123 Meta information generation unit 124 Vector Generation Unit 125 Clustering Department 126 Word Importance Information Update Unit 131 Reception 132 Search Section 133 Related diagram data generation unit 134 Output section 141 Information storage section [Prior art documents] [Patent documents]

[0185] [Patent Document 1] JP 2024-6420 A

Claims

1. a divided data generation unit that generates a plurality of divided data by dividing the contents of a plurality of document files; a search unit that searches the plurality of document files based on a search query input by a user; a related diagram data generating unit that generates related diagram data showing the relationship between the selected document file and other document files based on a portion of divided data that has a relatively high relevance to the search query among a plurality of divided data related to a selected document file selected by the user from the searched document files; An information processing device comprising:

2. the related diagram data generation unit calculates a similarity between one of the divided data as the part of the divided data and each of the plurality of divided data related to the other document file, and generates the related diagram data indicating a relationship between the other document file and the divided data having a relatively high similarity.

2. The information processing apparatus according to claim 1, wherein:

3. the related diagram data generation unit calculates a similarity between a cluster to which some of the divided data belongs, among a plurality of clusters obtained by clustering a plurality of divided data related to the selected document file, and each of a plurality of divided data related to the other document files, and generates the related diagram data indicating a relationship between the other document files and the selected document file for the divided data having a relatively high similarity; 2. The information processing apparatus according to claim 1, wherein:

4. the relationship diagram data generation unit calculates a similarity between a plurality of consecutive divided data sets including one divided data set having a relatively high relevance to the search query and each of a plurality of divided data sets related to the other document file, and generates the relationship diagram data indicating the relevance between the other document file and the divided data set having a relatively high similarity, and the selected document file; 2. The information processing apparatus according to claim 1, wherein:

5. the related diagram data generation unit calculates the similarity for each cluster obtained by clustering the plurality of divided data related to the other document file for each of the other document files, and generates the related diagram data indicating the relationship between the other document file and the selected document file for the other document file related to a cluster having a relatively high similarity; 5. The information processing apparatus according to claim 2, wherein the information processing apparatus is a computer.

6. a vector generation unit that generates, for each of the divided data, a divided data vector that indicates semantic features of the divided data; and the search unit searches the plurality of document files for document files associated with the divided data vectors that have a relatively high similarity to a search query vector that indicates semantic features of the search query; the part of the split data is split data relating to a split data vector that has a relatively high similarity to the search query vector among the plurality of split data relating to the selected document file; 2. The information processing apparatus according to claim 1, wherein:

7. the relationship diagram data generation unit calculates a similarity between a divided data vector relating to the part of the divided data and each divided data vector of the plurality of divided data relating to the other document file, classifies the divided data vectors of the plurality of divided data relating to the other document file having a relatively high similarity into a plurality of clusters by clustering, and generates the relationship diagram data showing the relationship between the selected document file and the clusters.

2. The information processing apparatus according to claim 1, wherein:

8. the related chart data generating unit generates the related chart data so as to include information indicating that the related chart data is based on the part of the divided data.

2. The information processing apparatus according to claim 1, wherein:

9. The information indicating that the information is based on the partial divided data is an icon.

9. The information processing apparatus according to claim 8,

10. the relationship diagram data generation unit generates the relationship diagram data so as to include an icon indicating that the similarity has been calculated using the cluster related to the other document file, the icon being different from an icon indicating that the similarity is calculated based on the part of the divided data.

6. The information processing apparatus according to claim 5,

11. the related diagram data generation unit is capable of selecting, when calculating the similarity between the selected document file and the other document file, a method of calculating the similarity for each cluster obtained by clustering a plurality of divided data related to the other document file for each of the other document files, or a method of calculating the similarity using a portion of the divided data of the other document file; 6. The information processing apparatus according to claim 5,

12. further comprising an output unit that displays the related chart data generated by the related chart data generation unit on a terminal.

2. The information processing apparatus according to claim 1, wherein:

13. the output unit causes the terminal to display the related diagram data showing the contents of the part of the divided data.

13. The information processing apparatus according to claim 12.

14. a divided data generation unit that generates a plurality of divided data by dividing the contents of a plurality of document files; a search unit that searches the plurality of document files based on a search query input by a user; a related diagram data generating unit that generates related diagram data showing the relationship between the selected document file and other document files based on a portion of divided data that has a relatively high relevance to the search query among a plurality of divided data related to a selected document file selected by the user from the searched document files; an information processing device having a terminal that displays a relational diagram based on the relational diagram data; An information processing system comprising:

15. a divided data generation procedure for generating a plurality of divided data by dividing the contents of a plurality of document files; a search step of searching the plurality of document files based on a search query input by a user; a related diagram data generating step of generating related diagram data showing the relationship between the selected document file and other document files based on a portion of divided data that has a relatively high relevance to the search query among a plurality of divided data related to a selected document file selected by the user from the searched document files; An information processing method characterized by being executed by a computer.

16. a divided data generation procedure for generating a plurality of divided data by dividing the contents of a plurality of document files; a search step of searching the plurality of document files based on a search query input by a user; a related diagram data generating step of generating related diagram data showing the relationship between the selected document file and other document files based on a portion of divided data that has a relatively high relevance to the search query among a plurality of divided data related to a selected document file selected by the user from the searched document files; A program that causes a computer to execute the following.

Images