Document search program, document search device, and document search method

Abstract

To provide a document retrieval apparatus, a document retrieval method and a document retrieval program, in document retrieval using retrieval keywords, which can provide the candidates of suitable retrieval keywords.SOLUTION: A document retrieval program according to an embodiment causes a computer to realize a document retrieval function, a word group creation function and a word group selection function. The document retrieval function retrieves a document related to a retrieval text inputted by a user in a document database stored with plural documents. The word group creation function creates plural word groups which classify suggested words extracted from the retrieved document using information about a document structure or classification information based on the meaning of words. The word group selection function selects the word group to be provided to the user from among the plural word groups.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a document search program, a document search device, and a document search method. [Background technology]

[0002] A document search system is a system that accepts search text such as search keywords and questions as inquiries from users, and outputs documents by referencing a document database (DB). Document search systems are provided on the Internet, and based on the search keywords and questions entered by the user, search for and output relevant documents from among the documents stored in the document database.

[0003] Conventional document search systems automatically collect documents published on the Internet and store, for each document, the words that appear in the document and the probability of each word appearing in the document. When a user makes a query, the system extracts documents from the stored set of documents in descending order of the probability of occurrence of words contained in the received input sentence, and outputs the extracted documents, as well as sentences and paragraphs that contain the words. Another known method performs morphological analysis on the natural language received as input sentences, identifies keywords contained in the input sentences, and automatically creates search queries.

[0004] However, in document search systems like those described above, users must think for themselves about appropriate keywords and phrases to search for documents containing the desired information. Therefore, some methods assist users in entering keywords by recommending words extracted from documents in search results as search keyword candidates. By recommending keywords that are likely to contain the desired information, users no longer need to think for themselves about appropriate keywords and phrases to search for documents containing the desired information, thereby enabling them to efficiently find the information they are looking for. These search keyword candidates are also called suggested words.

[0005] These search keyword candidates are selected from all sentences in a document. A document also contains various document components, such as titles, chapters, and paragraphs. Therefore, a variety of keywords selected from various document components are recommended as search keywords. This makes it difficult for users to compare search keywords and select keywords that will lead to documents containing the desired information. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] Japanese Patent Publication No. 2020-123131 Summary of the Invention [Problem to be solved by the invention]

[0007] The problem to be solved by the present invention is to provide a document retrieval device, a document retrieval method, and a document retrieval program that can present appropriate search keyword candidates in a document retrieval using a search keyword. [Means for solving the problem]

[0008] To solve these problems, a document search program according to an embodiment of the present invention causes a computer to implement a document search function, a word group generation function, and a word group selection function. The document search function searches a document database storing multiple documents for documents related to search text entered by a user. The word group generation function uses information about document structure or classification information based on word meanings to generate multiple word groups by classifying suggested words extracted from the searched documents. The word group selection function selects a word group to present to the user from the multiple word groups. [Brief explanation of the drawings]

[0009] [Figure 1]FIG. 1 is a diagram showing an example of the configuration of a document search system according to a first embodiment. [Figure 2] FIG. 1 is a diagram showing an example of the configuration of a document search device according to a first embodiment. [Figure 3] 10 is a flowchart illustrating a processing procedure of a document search process according to the first embodiment. [Figure 4] FIG. 3 is a diagram showing an example of a data flow in document search processing according to the first embodiment. [Figure 5] FIG. 2 is a diagram showing an example of a document used as a search target in the first embodiment. [Figure 6] FIG. 4 is a diagram for explaining an example of step S303 in FIG. 3. [Figure 7] FIG. 4 is a diagram for explaining an example of step S304 in FIG. 3. [Figure 8] FIG. 10 is a diagram for explaining a method of presenting suggested words according to a modified example of the first embodiment. [Figure 9] FIG. 10 is a diagram showing an example of the configuration of a document search apparatus according to a second embodiment. [Figure 10] 10 is a flowchart illustrating a processing procedure of a document search process according to the second embodiment. [Figure 11] FIG. 10 is a diagram showing an example of a data flow in a document search process according to the second embodiment. [Figure 12] FIG. 11 is a diagram for explaining an example of step S1002 in FIG. [Figure 13] FIG. 11 is a diagram for explaining an example of steps S1003 and S1004 in FIG. [Figure 14] FIG. 11 is a diagram for explaining an example of thesaurus information used in step S1003 of FIG. [Figure 15] 11 is a diagram for explaining a modified example of the thesaurus information used in step S1003 of FIG. [Figure 16] FIG. 10 is a diagram showing an example of the configuration of a document search apparatus according to a third embodiment. [Figure 17] FIG. 11 is a diagram showing an example of a data flow in a document search process according to the third embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, embodiments of a document search program, a document search device, and a document search method will be described in detail with reference to the drawings. In the following description, components having substantially the same functions and configurations are designated by the same reference numerals, and redundant explanations will be given only when necessary.

[0011] (First embodiment) 1 is a diagram showing the configuration of a document retrieval system 1 including a document retrieval device 100 according to a first embodiment. The document retrieval system 1 is a computer network system that performs an interactive document retrieval, in which document data matching a query from a document database is selected based on the query from a user. As shown in FIG. 1, the document retrieval system 1 includes the document retrieval device 100, a document DB (document database) 200, and a client terminal 300.

[0012] The document search device 100 is connected to a document DB 200 and a client terminal 300 via a network or the like. The network is, for example, a LAN (Local Area Network). The connection to the network may be a wired connection or a wireless connection. Furthermore, the network is not limited to a LAN and may be the Internet, a public communication line, or the like.

[0013] The document DB 200 is a computer that maintains a database that stores multiple document data items to be searched. The document data may be, for example, in HTML or PDF format, but may also be in any other format. In the following description, "document data" will be simply referred to as "document." Each document is composed of multiple sentences, words, symbols, etc. The type of document to be searched may be a report, a manual, or any other type of document.

[0014] The client terminal 300 is a computer used by a user of the document retrieval system 1. The client terminal 300 has, as hardware, a processor, an input device, a display device, and a communication device, and functions as a user interface of the document retrieval system 1. For example, the client terminal 300 accepts input of text related to a document to be searched (hereinafter referred to as search text) input by the user via the input device. The client terminal 300 transmits the input search text to the document retrieval device 100.

[0015] The document search device 100 functions as a server device of the document search system 1. Specifically, the document search device 100 receives search text from the client terminal 300, searches for documents related to the search text from among the documents stored in the document DB 200 based on the received search text, and transmits the searched documents (hereinafter referred to as search documents) to the client terminal 300.

[0016] The document search device 100 also selects from the searched documents a number of search keywords to recommend to the user in order to further narrow down the searched documents, and transmits the extracted search keywords to the client terminal 300. The client terminal 300 receives the search keywords and document search results from the document search device 100 and displays them on a display device. In the following description, the search keywords presented to the user and candidate search keywords to be presented are referred to as suggested words.

[0017] Fig. 2 is a diagram showing an example of the configuration of the document retrieval device 100. As shown in Fig. 2, the document retrieval device 100 is a computer having a processing circuit 11, a storage device 12, an input device 13, a communication device 14, and a display device 15. Data communication between the processing circuit 11, the storage device 12, the input device 13, the communication device 14, and the display device 15 is performed via a bus.

[0018] The processing circuit 11 includes a processor such as a CPU (Central Processing Unit) and a memory such as RAM (Random Access Memory). The processing circuit 11 includes a document search unit 111, a block division unit 112, a word group generation unit 113, a word group selection unit 114, and a suggestion word presentation unit 115. The processing circuit 11 executes a document search program to realize the document search function, block division function, word group generation function, word group selection function, and suggestion word presentation function of the above units. The document search program is stored in a non-transitory computer-readable recording medium such as the storage device 12. The document search program may be implemented as a single program that describes all the functions of the above units, or as multiple modules divided into several functional units. Furthermore, the above units may be implemented by an integrated circuit such as an application-specific integrated circuit (ASIC). In this case, the units may be implemented on a single integrated circuit or individually on multiple integrated circuits.

[0019] The storage device 12 is configured by a ROM (Read Only Memory), an HDD (Hard Disk Drive), an SSD (Solid State Drive), an integrated circuit storage device, etc. The storage device 12 stores a document search program and the like.

[0020] Various commands are input from an operator via the input device 13. A keyboard, a mouse, various switches, a touchpad, a touch panel display, etc. can be used as the input device 13. An output signal from the input device 13 is supplied to the processing circuit 11.

[0021] The communication device 14 is an interface for performing data communication with an external device connected to the document search device 100 via a network. As an example, the communication device 14 performs data communication with the document DB 200 and the client terminal 300.

[0022] The display device 15 displays various information. As the display device 15, a CRT (Cathode-Ray Tube) display, a liquid crystal display, an organic EL (Electro Luminescence) display, an LED (Light-Emitting Diode) display, a plasma display, or any other display known in the art can be appropriately used. The display device 15 may also be a projector.

[0023] Next, the functions performed by each unit of the processing circuit 11 will be described in detail. The document search unit 111 searches documents stored in the document DB 200 for documents related to search text input by the user, and acquires the searched documents from the document DB 200. The search text is input by the user at the client terminal 300. The search text may be input as natural language such as a question, or as words such as search words or suggested words (described later). The words may be input as a single word or as a word string containing multiple words. The document search unit 111 extracts words from the search text received from the client terminal 300, searches the document DB 200 for documents containing the extracted words, and acquires the searched documents as search documents. The document search unit 111 outputs the acquired search documents as candidate documents to be provided to the user.

[0024] The block division unit 112 divides each retrieved document into multiple document blocks based on information about the document structure. For example, the block division unit 112 divides each retrieved document into multiple document blocks by classifying each sentence included in the retrieved document into one of multiple document blocks according to the document structure. The information about the document structure includes, for example, HTML tags attached to the document, sentences included in each paragraph number, and the order of each sentence in the document. The block division unit 112 outputs block information about the document blocks including the division results of each document as information about the document structure.

[0025] The word group generation unit 113 generates multiple word groups by classifying suggested words extracted from search documents. Specifically, the word group generation unit 113 extracts suggested words to be used for narrowing down documents for each document block and generates word groups composed of the suggested words for each document block. In this case, the word group generation unit 113 generates word groups composed of suggested words extracted from the same type of document block in each document for each document block. Suggested words are words used for narrowing down documents and can also be called recommended search words effective for narrowing down document searches. Suggested words are typically single nouns, but may also be adjectives, verbs, etc., or sentences containing multiple words. Suggested words can be extracted using, for example, known natural language processing. Suggested words extracted from search documents are used as candidates for suggested words to be presented to the user. The word group generation unit 113 may also be called a word group extraction unit.

[0026] When extracting suggested words, the word group generation unit 113 extracts all words contained in each document as suggested words, for example, by performing natural language analysis on the documents. Furthermore, when generating word groups, the word group generation unit 113 extracts suggested words for each document block, for example, and generates word groups for each document block by collecting suggested words extracted from the same document block from each search document. The word group generation unit 113 outputs the generated word groups and sets of suggested words belonging to the word groups as information about the word groups.

[0027] Each word group contains suggested words from documents extracted from the same document block. Multiple suggested words may be extracted from one document in each word group. Suggested words belonging to the document block word group are not extracted from documents that do not contain that document block. Therefore, a word group may be composed of suggested words extracted from a portion of the retrieved documents.

[0028] The word group selection unit 114 selects a word group to be presented to the user from among multiple word groups. At this time, the word group selection unit 114 selects a word group to be presented to the user based on a score indicating the document narrowing-down performance of the suggestion word. Specifically, the word group selection unit 114 calculates a score for each word group (hereinafter referred to as a group score) based on a score indicating the document narrowing-down performance of each suggestion word belonging to each word group (hereinafter referred to as a word score), compares the group scores of each word group, and selects a word group (hereinafter referred to as a selected group) to be presented to the user. At this time, the word group selection unit 114 selects a word group with a high word score, thereby selecting a word group composed of suggestion words with high document narrowing-down performance. The number of selected word groups may be one or more.

[0029] The word score, which indicates document narrowing performance, is an index indicating whether a word is an important word that characterizes a document. Examples of word scores that can be used include the number of occurrences in a document, entropy based on the number of occurrences, and tf-idf. The word score can be calculated using all documents stored in the document DB 200, or related documents extracted in response to user input. The word score for each suggested word can be calculated and stored in advance, or calculated when the group score is calculated. Examples of group scores that can be used include the sum, maximum value, median, and average value of the word scores for each suggested word belonging to a word group. The suggested words used to calculate the group score can also be limited. For example, a predetermined number of suggested words can be extracted from each word group in descending order of word score, and the group score can be calculated using only the word scores of the extracted suggested words. The group score can be used as an index indicating the effectiveness of the suggested words belonging to that word group in document narrowing.

[0030] The suggestion word presentation unit 115 presents the suggestion words belonging to the selected word group to the user. Specifically, the suggestion word presentation unit 115 displays the suggestion words belonging to the selection group selected by the word group selection unit 114 on the display device 15 or the display device of the client terminal 300. At this time, only a predetermined number of suggestion words belonging to the selection group may be displayed in descending order of word score from the suggestion words belonging to the selection group.

[0031] Of the suggested words contained in the searched document, only the suggested words that belong to the selected word group are displayed on the display device of the client terminal 300. Each word group is generated for each document block classified according to its position in the document. Therefore, for one word group, suggested words with the same position in the document are extracted from each document. Therefore, only suggested words with the same position in the document are presented to the user.

[0032] In addition, the suggested words presented to the user are selected from word groups consisting of suggested words with high document narrowing performance based on the word scores of the suggested words to which they belong, so that only important words that characterize the document are presented to the user.

[0033] (Document search processing) Next, the operation of the document search process executed by the document search device 100 will be described. The document search process is a process of presenting suggested words to narrow down the documents that the user wants to acquire. The document search device 100 starts the document search process based on receiving a question sentence or search words input by the user from the client terminal 300. FIG. 3 is a flowchart showing an example of the procedure of the document search process. FIG. 4 is a diagram showing an example of the data flow in the document search process. Note that the processing procedures in each process described below are merely examples, and each process can be modified as appropriate as possible. Furthermore, steps in the processing procedures described below can be omitted, replaced, or added as appropriate depending on the embodiment.

[0034] Here, as an example, we will explain the case where a single word is entered as a search word, and suggested words are presented to help the user narrow down the documents they need from HTML format documents (hereinafter referred to as procedural documents) that explain various procedures.

[0035] (Step S301) In the document search process, first, the document search unit 111 searches documents stored in the document DB 200 for documents related to the search word acquired from the client terminal 300. At this time, the document search unit 111 searches the document DB 200 for all documents that include the search word, and acquires the retrieved documents as search documents from the document DB 200. The document search unit 111 outputs the acquired search documents to the block division unit 112 as document search results.

[0036] (Step S302) Next, the block division unit 112 divides each document included in the search document into a plurality of document blocks. Fig. 5 is a diagram showing an example of the document structure of a procedural document 500 in HTML format. The procedural document 500 shown in Fig. 5 can be divided into a first document block 510 in which the "procedure name" is written, a second document block 520 in which the "paragraph names for each item related to the procedure" are written, and a third document block 530 in which the "items" are written. Each sentence included in the procedural document 500 in Fig. 5 has an HTML tag indicating that it is classified as a "procedure name": <h1> HTML tag indicating that the item is classified as a "paragraph name for each item related to the procedure"< / h1> <h2>、 , and the HTML tag indicating that it is classified as a "bullet item" The block division unit 112 uses HTML tags as document structure information to classify each sentence included in the retrieved document into one of document blocks: "procedure name," "paragraph name for each item related to the procedure," and "item." Using the classification results for each sentence, the block division unit 112 divides each document into document blocks corresponding to "procedure name," "paragraph name for each item related to the procedure," and "item." The block division unit 112 outputs the division results for each document to the word group generation unit 113 as block information.

[0037] (Step S303) Next, the word group generation unit 113 generates word groups for each document block using the results of the document block division. FIG. 6 is a schematic diagram showing how word groups are generated from multiple procedural documents 500. The word group generation unit 113 extracts suggested words from the first document block 510 of each procedural document 500 and generates a first word group 610 using the extracted suggested words. Similarly, the word group generation unit 113 generates a second word group 620 consisting of suggested words extracted from the second document block 520 of each procedural document 500 and a third word group 630 consisting of suggested words extracted from the third document block 530 of each procedural document 500. Each word group 610, 620, and 630 contains only suggested words extracted from the same document block. The word group generation unit 113 outputs the generated word groups 610, 620, and 630 to the word group selection unit 114, along with the suggested words belonging to each word group.

[0038] (Step S304) Next, the word group selection unit 114 selects a word group to present to the user from the generated word groups 610, 620, and 630. Fig. 6 is a schematic diagram showing how a word group is selected from the word groups 610, 620, and 630. In this case, the word group selection unit 114 first calculates the entropy of each suggested word as a word score. The numerical value next to each suggested word in Fig. 7 indicates the entropy of each suggested word.

[0039] Next, the word group selection unit 114 calculates the sum of the entropy values ​​of the suggested words belonging to each word group as the group score for each word group 610, 620, 630. The word group selection unit 114 compares the group scores of each word group 610, 620, 630, and selects the first word group 610 with the highest group score as the word group to present to the user. Instead of the sum of the word scores, the maximum, median, average, etc. of the word scores may be used as the group score. The word group selection unit 114 outputs the selected first word group 610 to the suggested word presentation unit 115 as the word group selection result.

[0040] (Step S305) Next, the suggestion word presentation unit 115 transmits all of the suggestion words included in the selected first word group 610 to the client terminal 300, and displays the transmitted suggestion words on the display device of the client terminal 300, thereby presenting the suggestion words to the user. At this time, in addition to the suggestion words included in the first word group 610, a list of searched documents is simultaneously displayed.

[0041] Alternatively, only a predetermined number of suggested words may be presented to the user. In this case, the suggested word presentation unit 115, for example, refers to the word scores of the suggested words included in the first word group 610, selects a predetermined number of suggested words in descending order of word score, and presents the selected suggested words to the user.

[0042] The user can select a suggested word related to the document he or she wants to search from the suggested words presented.

[0043] When the user selects one of the suggested words (step S306-YES), the document search device 100 executes the processes of steps S301 to S305 again. At this time, the document search device 100 uses the selected suggested word as search text to search for documents containing the selected suggested word from among the searched documents, and narrows down the documents again. Thereafter, the document search device 100 divides the narrowed down searched documents into document blocks, generates word groups for each document block, selects word groups to be presented again, and presents the suggested words included in the presented word groups to the user. This updates the suggested words presented to the user.

[0044] The user repeats the operation of selecting one suggested word from the multiple suggested words presented until the desired document is found. This causes the processes of steps S301 to S306 to be repeatedly executed, narrowing down the documents and presenting suggested words repeatedly. When the user finds the desired document (step S306-NO), the document search process ends.

[0045] The effects of the document search device 100 according to this embodiment will be described below.

[0046] The document search device 100 according to this embodiment includes a document search unit 111, a block division unit 112, a word group generation unit 113, and a word group selection unit 114. The document search unit 111 searches a document DB 200, which stores multiple documents, for documents related to search text entered by a user. Specifically, the document search unit 111 searches the document DB 200 for documents that contain words included in the search text. The word group generation unit 113 generates multiple word groups by classifying suggestion words extracted from the searched documents. The word group selection unit 114 selects a word group to present to the user from the multiple word groups.

[0047] The document search device 100 according to this embodiment further includes a block division unit 112. The block division unit 112 divides each search document into multiple document blocks. For example, the block division unit 112 classifies each sentence included in the search document into multiple document blocks based on the document structure. Examples of document structure include HTML tags, the order of sentences in the document, and paragraph numbers of sentences. The block division unit 112 divides each document into multiple document blocks by classifying each sentence included in the document into one of multiple document blocks. The word group generation unit 113 extracts suggested words to be used for document narrowing from the search document based on information about the document structure, and generates, for each document block, word groups composed of suggested words extracted from the same type of document block in each document. The word group selection unit 114 selects word groups to present to the user based on word scores indicating the document narrowing performance of the suggested words. At this time, the word group selection unit 114 calculates a group score for each word group using the word scores of the suggested words belonging to each word group, and selects a word group to present by comparing the group scores.

[0048] The word score can be, for example, an entropy score based on the number of occurrences of a word, the number of occurrences in a document, tf-idf, or other numerical values ​​that indicate the degree to which a word is important in characterizing a document. The group score can be, for example, the sum, median, average, or maximum value of the word scores.

[0049] The document search device 100 further includes a suggestion word presentation unit 115. The suggestion word presentation unit 115 presents to the user suggestion words that belong to the selected word group.

[0050] With the above configuration, the document search device 100 according to this embodiment classifies the suggested word candidates extracted from each document into document blocks according to the document structure, thereby dividing the suggested word candidates into document blocks according to their position in the document. Then, by recommending only words contained in specific document blocks as suggested words to be used in a narrowed search, only words with similar positions in the document can be presented to the user as suggested words. When selecting suggested words to narrow down documents, multiple suggested words with similar positions in the document are presented. This allows the user to narrow down documents by comparing important words that characterize the documents, thereby more efficiently searching for documents containing desired information.

[0051] For example, as shown in Figures 5-7, when a first word group 610 composed of suggested words belonging to a first document block 510 is selected, a list of suggested words classified as "procedure name" is presented to the user. Words with the same position in the document are presented as search word options, allowing the user to appropriately determine which words to select to find a document containing the desired information. Suggested words may also be classified into multiple document blocks based on the order of sentences in the document or the paragraph numbers of the sentences. Sentences or paragraphs with the same order in the document are similarly positioned in the document, so by presenting only suggested words belonging to the selected document block, it is possible to present only words with similar positions in the document to the user.

[0052] Furthermore, the word group selection unit 114 selects word groups made up of suggested words with high document narrowing performance based on word scores that indicate document narrowing performance. This allows words that are highly important in a document and useful for distinguishing it from other documents to be presented to the user. Since the user can narrow down documents by comparing important words that characterize the documents, they can more efficiently find documents that contain the information they are looking for.

[0053] (Modification of the first embodiment) The number of word groups presented to the user may be one or more. When presenting multiple word groups to the user, the word group selection unit 114 selects, for example, the top two word groups with the highest group scores, and the suggestion word presentation unit 115 presents the user with suggested words belonging to the selected multiple word groups. In this case, the suggestion word presentation unit 115 may display information about the word group to which each suggested word belongs, in association with the suggested word. For example, as shown in FIG. 8, the name of the word group to which the suggested word belongs may be displayed in association with each suggested word. When suggestion words belonging to the selected word group are presented to the user, information about the word group to which the suggested word belongs is presented so that the user can understand which group each suggested word belongs to. This allows the user to select keywords to search for documents containing the information they are looking for by comparing suggested words belonging to the same word group. Furthermore, only when the number of suggested words belonging to the selected word group is small, multiple word groups may be selected to ensure the number of suggested words presented to the user. Even when one word group is presented to the user, information such as the name of the selected word group may be displayed along with the suggested words. By understanding the selected word group, the user can understand the position of the suggested words in the document.

[0054] Instead of displaying the name of the word group, the suggested word may be displayed in a manner that distinguishes the word group to which it belongs by changing the color, font, and display position of the letters depending on the word group to which each suggested word belongs.

[0055] (Second embodiment) A second embodiment will be described. This embodiment is a modification of the configuration of the first embodiment as follows. Descriptions of the same configuration, operation, and effects as those of the first embodiment will be omitted. In this embodiment, multiple word groups are generated by classifying suggested words according to the meaning of the words.

[0056] 9 is a diagram showing an example of the configuration of the document search device 100 according to this embodiment. As shown in Fig. 9, the processing circuitry 11 includes a suggest word extraction unit 116 and a word group generation unit 117, instead of the block division unit 112 and the word group generation unit 113 shown in Fig. 2.

[0057] The suggestion word extraction unit 116 extracts suggestion words to be used for narrowing down documents from the searched document. For example, the suggestion word extraction unit 116 extracts a plurality of suggestion words from the entire searched document by performing known natural language processing on each sentence included in the searched document.

[0058] The word group generation unit 117 generates a plurality of word groups by classifying the suggested words extracted from the search document. The word group generation unit 117 of this embodiment uses classification information based on word meanings to classify each extracted suggested word into a plurality of categories and generate a word group for each category. The classification information based on word meanings is information that defines, for example, the meanings of words, the hierarchical relationships of words, the hierarchical relationships of words, the hierarchical relationships of words in the conceptual hierarchy, the hierarchical relationships based on the inclusion relationships of words, and the similarity relationships of words. The classification information based on word meanings can be, for example, thesaurus information. In this case, a general Japanese thesaurus may be used, or a thesaurus specific to the industry to which the document to be searched belongs may be used. Using an industry-specific thesaurus can improve classification accuracy. The classification information based on word meanings may be stored in advance in the storage device 12 or obtained from an external system.

[0059] (Document search processing) Next, the operation of the document search process executed by the document search device 100 according to this embodiment will be described. FIG. 10 is a flowchart showing an example of the procedure of the document search process. FIG. 11 is a diagram showing an example of the data flow in the document search process. Note that the processing procedures in each process described below are merely examples, and each process can be modified as appropriate as possible. Furthermore, steps in the processing procedures described below can be omitted, replaced, or added as appropriate depending on the embodiment.

[0060] Here, as an example, a case will be described in which suggested words are presented to narrow down the documents the user needs from documents that explain various insurance policies in the insurance industry (hereinafter referred to as insurance documents). (Step S1001) Next, the document search unit 111 searches documents stored in the document DB 200 for documents related to the search word acquired from the client terminal 300, in the same manner as in step S301 in the first embodiment.

[0061] (Step S1002) Next, the suggestion word extraction unit 116 extracts suggestion words from each searched document. Fig. 12 is a diagram showing an example of an insurance document 1200. Fig. 12 is also a schematic diagram showing how suggestion words are extracted from a plurality of insurance documents 1200. As shown in Fig. 12, the suggestion word extraction unit 116 extracts suggestion words from all the searched insurance documents 1200 and outputs the extracted suggestion word list 1210 to the word group generation unit 117.

[0062] (Step S1003) Next, the word group generation unit 117 generates multiple word groups using the meanings and inclusion relationships of the suggested words as classification information based on the meanings of the words. FIG. 13 is a schematic diagram showing how multiple word groups are generated by classifying each suggested word included in the extracted suggested word list 1210. FIG. 14 is a diagram showing an example of a thesaurus for the insurance industry. In FIG. 13, the word group generation unit 117 classifies each suggested word included in the suggested word list 1210 into one of "coverage target," "procedure content," and "service content" using the insurance industry thesaurus shown in FIG. 14. Then, the word group generation unit 117 generates a first word group 1310 to which the word used as "coverage target" belongs, a second word group 1320 to which the word used as "procedure content" belongs, and a third word group 1330 to which the word used as "service content" belongs. For example, if the word "automobile" is classified as "coverage" in an insurance industry thesaurus, the suggested word "automobile" is determined to belong to the first word group 1310, as shown in Fig. 13. The word group generation unit 113 outputs the generated word groups 1310, 1320, and 1330 to the word group selection unit 114, together with the suggested words belonging to each word group. 13 and 14 have described an example in which a thesaurus for the insurance industry is used as classification information based on the meaning of words, but each extracted suggested word may be classified using a general Japanese thesaurus as shown in Fig. 15. In this case, the word group generation unit 117 classifies each suggested word into categories such as "medical care," "biology," and "vehicles," and generates word groups to which the classified words belong for each category.

[0063] (Step S1004) Next, the word group selection unit 114 calculates a group score for each word group using the word score of each suggested word, similar to the processing in step S304 of the first embodiment, and selects a word group to present to the user by comparing the group scores of each word group.

[0064] (Step S1005) Next, the suggestion word presentation unit 115 presents to the user suggestion words included in the word group selected as the word group, in the same manner as in the process of step S305 in the first embodiment.

[0065] Furthermore, similarly to the first embodiment, the user can repeat the operation of selecting one suggested word from the plurality of suggested words presented until the user finds the desired document.

[0066] The effects of the document search device 100 according to this embodiment will be described below.

[0067] The document search device 100 according to this embodiment includes a document search unit 111, a word group selection unit 114, a suggestion word extraction unit 116, and a word group generation unit 117. The suggestion word extraction unit 116 extracts suggestion words to be used to narrow down documents from all searched documents. The word group generation unit 117, like the word group generation unit 113 of the first embodiment, generates multiple word groups by classifying the suggestion words extracted from the searched documents. In this case, the word group generation unit 117 generates multiple word groups using classification information based on word meanings. The classification information includes, for example, word meanings, word hierarchical relationships, word hierarchical relationships in a conceptual hierarchy, or hierarchical relationships based on word inclusion relationships. The word group generation unit 117 classifies the extracted suggestion words into multiple categories using, for example, thesaurus information containing the classification information, and generates multiple word groups for each classification result. As in the first embodiment, the document search device 100 further includes a suggested word presentation unit 115 that presents suggested words belonging to the selected word group to the user.

[0068] With the above configuration, in this embodiment too, candidate suggestion words extracted from each document are classified into multiple word groups according to their position in the document, and only words belonging to specific word groups are recommended as suggested words to be used in narrowed-down searches, so that, as in the first embodiment, only words that are similarly positioned in the document can be presented to the user as suggested words.

[0069] For example, as shown in Figures 12 and 13, when the first word group 1310 is selected, a list of suggested words categorized as "covered by insurance" is presented to the user. Words with the same position in the document are presented as search word options, allowing the user to appropriately determine which words to select to find a document containing the desired information.

[0070] Furthermore, as in the first embodiment, a word group composed of words in a document that have high document narrowing performance is selected, so that the user can narrow down the documents using suggested words that represent the characteristics of the document, and can more efficiently find documents that contain the information the user is looking for.

[0071] (Modification of the second embodiment) When suggestion words belonging to a selected word group are presented to a user, thesaurus information may be used to combine multiple words classified into the same meaning or type into a single word. For example, among the suggestion words belonging to the first word group 1310 in FIG. 13, the words "dog" and "cat" may be combined into the word "animal" and presented. In this case, for example, the sum or average of the word scores of "dog" and "cat" is used as the word score for "animal." When selecting and presenting a predetermined number of suggestion words with high word scores from the suggestion words belonging to the selected word group, combining and presenting words with similar meanings makes it possible to present a wider variety of words as options to the user.

[0072] (Third embodiment) A third embodiment will now be described. This embodiment is a modification of the configurations of the first and second embodiments as follows. Descriptions of configurations, operations, and effects similar to those of the first and second embodiments will be omitted. In this embodiment, multiple word groups are generated based on the document structure, and further multiple word groups are generated based on the meanings of words.

[0073] Fig. 16 is a diagram showing an example of the configuration of the document search device 100 according to this embodiment. As shown in Fig. 16, the processing circuitry 11 includes a document search unit 111, a block division unit 112, a word group generation unit 113, a word group selection unit 114, and a suggestion word presentation unit 115 described in the first embodiment, as well as a suggestion word extraction unit 116 and a word group generation unit 117 described in the first embodiment.

[0074] 17 is a diagram showing an example of a data flow in the document search process executed by the document search device 100 according to this embodiment. In this embodiment, the document search unit 111 outputs the document search results to both the block division unit 112 and the suggestion word extraction unit 116.

[0075] As in the first embodiment, the block division unit 112 divides each search document into multiple document blocks, and outputs the division results as block information to the word group generation unit 113. The word group generation unit 113 generates multiple word groups for each document block, and outputs the generated word groups to the word group selection unit 114.

[0076] As in the second embodiment, the suggestion word extraction unit 116 extracts suggestion words from each search document and outputs the extracted suggestion words to the word group generation unit 117. The word group generation unit 117 generates a plurality of word groups by classifying the suggestion words based on the meanings of the words, and outputs the generated word groups to the word group selection unit 114.

[0077] The word group selection unit 114 acquires multiple word groups from each of the word group generation units 113 and 117, and selects a word group to present to the user from all of the acquired word groups. In this case, the word group selection unit 114 calculates a group score for all of the acquired word groups using the word scores of the suggested words, and selects the word group with the highest group score from all of the word groups as the word group to present. Note that the word group with the highest group score may be selected, or multiple word groups may be selected in descending order of group score.

[0078] With the above configuration, the document search device 100 of this embodiment can select word groups that are more useful for document narrowing down by selecting word groups with high group scores that represent the document narrowing down performance within a document from both word groups classified based on document structure and word groups classified according to classifications based on word meanings.

[0079] The word group selection unit 114 may randomly select word groups to present without using the word scores and group scores. Even in this case, only words belonging to a specific word group are presented as suggested words, so the user can easily compare suggested words that are similarly positioned in the document and appropriately determine which words to select to find a document that contains the information the user is looking for.

[0080] Thus, according to any of the above-described embodiments, it is possible to provide a document retrieval program, a document retrieval device, and a document retrieval method that can present appropriate search keyword candidates in a document retrieval using a search keyword.

[0081] The present invention is not limited to the above-described embodiments, and the components can be modified and embodied in practice without departing from the spirit of the invention. Furthermore, various inventions can be created by appropriately combining multiple components disclosed in the above-described embodiments. For example, some components may be omitted from all the components shown in the embodiments. Furthermore, components from different embodiments may be appropriately combined. [Explanation of symbols]

[0082] 1...document search system, 100...document search device, 200...document DB, 300...client terminal, 11...processing circuit, 111...document search unit, 112...block division unit, 113, 117...word group generation unit, 114...word group selection unit, 115...suggested word presentation unit, 116...suggested word extraction unit, 12...storage device, 13...input device, 14...communication device, 15...display device, 500, 1200...documents, 510, 520, 530...document blocks, 610, 620, 630, 1310, 1320, 1330...word groups, 1210...suggested word list. < / h2>

Claims

1. On the computer, A document search function that searches a document database containing multiple documents for documents containing words included in the search text entered by the user; a block division function for dividing each of the retrieved documents into a plurality of document blocks based on information about the document structure; a word group generation function that extracts suggested words used to narrow down documents for each document block and generates word groups made up of the suggested words for each document block; a word group selection function for selecting a word group to be presented to a user based on a score indicating the document narrowing performance of the suggested words; A document search program to achieve this.

2. The information about the document structure includes HTML tags, the order of sentences, or paragraph numbers of sentences.

2. The document retrieval program according to claim 1.

3. the word group selection function calculates the sum, average, maximum, or median of the score for each of the word groups, and selects a word group to be presented to the user based on the calculation result; 2. The document retrieval program according to claim 1.

4. the word group selection function selects one word group as a word group to be presented to the user; 2. The document retrieval program according to claim 1.

5. the word group selection function selects a plurality of word groups as word groups to be presented to the user; 2. The document retrieval program according to claim 1.

6. To further realize a suggestion word presentation function that presents to the user suggestion words that belong to the selected word group, 2. The document retrieval program according to claim 1.

7. the suggestion word presentation function presents to the user information indicating the word group to which the suggestion word belongs, in addition to the suggestion words belonging to the selected word group; 7. The document retrieval program according to claim 6.

8. a document search unit that searches a document database storing a plurality of documents for documents related to a search text input by a user; a block dividing unit that divides each of the retrieved documents into a plurality of document blocks based on information about the document structure; a word group generating unit that extracts suggested words to be used for narrowing down documents for each of the document blocks and generates word groups made up of the suggested words for each of the document blocks; a word group selection unit that selects a word group to be presented to a user based on a score indicating the document narrowing performance of the suggested words; A document search device comprising:

9. Searching for documents related to a search text entered by a user from a document database storing a plurality of documents; Dividing each of the retrieved documents into a plurality of document blocks based on information about the document structure; extracting suggested words to be used for narrowing down documents for each of the document blocks, and generating word groups made up of the suggested words for each of the document blocks; selecting a word group to be presented to a user based on a score indicating document narrowing performance of the suggested words; A document retrieval method comprising:

Images