JP2009032180A - Text mining apparatus and text mining method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method for efficiently collecting documents by retrieval based on text mining and document quotation information. <P>SOLUTION: This system comprises: a condition input processing part for providing a bio-database (document database) for a text mining target and an input screen for urging a user to input a keyword necessary for document retrieval and accepting the keyword input from the input screen; a text mining processing part for executing text mining retrieval of the document database on the basis of a condition accepted by the condition input part; a document quotation relation analysis processing part for analyzing the quotation relation of documents extracted by the text mining retrieval; and a result output processing part for displaying document names which are the results of the text mining retrieval and their quotation relation information like a bird's-eye view. The system is also provided with a means for storing the database (document database) necessary for evaluation in the latest state. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a text mining technology, and in searching for academic papers in the life science field, searches for a group of papers to be acquired based on the citation relationship between the text mining technology and the documents based on published document information. Regarding technology.

  The results of experiments in the field of life science are mostly text documents in the form of biodatabases, such as base sequence data, protein amino acid sequence data, protein tertiary structure data, and interaction data between biomolecules. Saved and published in

  These data are accompanied by annotation (annotation) information and bibliographic information, and recently, necessary information from a large amount of data published using text mining technology for these text format data. The means to extract is becoming used.

  However, the enormous amount of data disclosed in the world is not only satisfying the enormous amount of data, but also because there are many data with different characteristics and properties, so only the first search result by the system is satisfactory. It is difficult to collect information, and the necessary amount of data is obtained by repeating screening multiple times at the discretion of the user. In particular, regarding the acquisition of document information, the work of collecting the necessary related documents by tracing the citation information of a certain document in the search result by keyword is mixed, so what is the finally collected document group? There is a reality that it is difficult to re-arrange what is collected by various conditions and methods. This situation is particularly likely to occur when data is collected using a wide variety of search services.

  It should be noted that the method of searching related documents by tracing the citation information of documents is significant in the field where many documents are published in a short period of time because published documents are cited from other documents in a short period of time. It is used as a method for searching for documents to be collected by analyzing documents that have obtained a large number of citations. In particular, the life science field including fields such as medicine and biology is a field in which a large number of documents are published in a short period of time, and is effective for collecting documents that use citation information of documents.

  Note that text mining is a technique for efficiently extracting useful knowledge and information by analyzing a large-scale text database from various viewpoints using a computer. Text mining combines multiple elemental technologies such as natural language processing and information visualization. By using text mining, you can select texts that contain the information you want, analyze the relationships between texts and the relationships between items described in the text, and obtain information that cannot be obtained by reading individual texts. It is expected to extract keywords that characterize a given sentence set. Such text mining is attracting attention for its application to customer requirement analysis, etc., but since large document databases such as MEDLINE can be freely used, research on text mining in the pharmaceutical field is becoming active. is there.

  As a document related to the text mining method, there is Patent Document 1 below.

JP 2001-318948 A

  Conventionally, as described above, a text mining search that directly extracts features from text information described in the contents of data to be collected, and a process of searching for other document information by tracing the citation relationship of a document There is a problem in that it is necessary to separately and separately manage the conditions, methods, and order in which the finally collected documents are collected.

  An object of the present invention is to reduce the burden of the above processing.

  According to an aspect of the present invention, a user interface for prompting a client terminal to input a keyword for document search is provided, and a condition for performing a text mining search process based on the input keyword The text mining condition input processing unit that accepts the text, the text mining processing unit that directly mines the text into the literature database based on the accepted condition, and the citation relation information of the literature extracted by the processing result of the text mining A document citation relationship table created by extracting only information, and analyzing based on a table in which direct citation relationships are arranged as pairs of documents, and the text mining processing and the document citation relationship Based on the analysis process, the number of cited references, the number of cited references, and the key A number of documents that were extracted by the text mining corresponding to over-de, a text mining apparatus is provided, characterized in that it comprises a display control unit for performing document-specific display control for displaying in distinction for each document.

  When the display control unit specifies a document, it is preferable to perform hierarchical display control for hierarchically displaying the citation relationship with respect to the cited document having a relationship in which the document is cited. Moreover, it is preferable that the said display control part performs the control which displays the said hierarchical display and the said literature display on the same screen, or switching.

  According to another aspect of the present invention, a user interface for prompting a client terminal to input a keyword for literature search is provided, and a text mining search process is performed based on the input keyword. Based on the accepted conditions, the step of text mining directly into the literature database based on the accepted conditions, and the citation relation information of the literature extracted by the processing result of text mining by extracting only the literature citation information The number of cited documents, the number of cited documents, and the number of documents extracted by text mining corresponding to the keyword based on the step of creating the document citation relationship table, the text mining process, and the document citation relationship analysis process And a step of distinguishing and displaying each document. Text mining method according to is provided.

  Moreover, a means for keeping a biodatabase (document database) necessary for evaluation in the latest state may be provided.

  The present invention may be a program for causing a computer to execute the above steps, or a computer-readable recording medium for recording the program.

  According to the present invention, when collecting documents related to a specific theme, using a document citation relationship table created by extracting only document citation information, together with document search by text mining, it is useful according to the number of cited documents. By searching for documents, useful documents can be collected. Therefore, it is possible to save time and labor for the document collection process. In addition, it is possible to easily perform a display for analyzing the citation relationship from the literature citation relationship table.

  Hereinafter, a text mining technique according to an embodiment of the present invention will be described with reference to the drawings.

  FIG. 1 is a functional block diagram showing a system configuration example including a text mining device according to an embodiment of the present invention. The system according to the present embodiment includes a client computer (hereinafter simply referred to as “client”) 3 for a user to input keywords and refer to evaluation results in order to use the system, and a server for text mining. A computer (hereinafter simply referred to as “server”) 1, a document database 5 that stores document information, and a document citation relationship index table 6 that can index document citation relationships are connected to each other by a network 2. .

  The server 1 provides an interface for prompting a user to input a keyword necessary for collecting documents, and a text mining condition receiving means for receiving a condition for performing a text mining search based on the keyword. As a condition input processing unit 9, a text mining processing unit 8 as means for text mining in the document database 5 according to the accepted conditions, and a document citation relation index table 6 in which the citation relations of documents extracted by the text mining result are stored. The document citation relationship analysis processing unit 7 as a means for analyzing the citation relationship of documents, the result output processing unit 10 as a means for evaluating and displaying these results in combination, and the document database necessary for the evaluation are updated. The document database update processing unit 4 is provided as means for maintaining the state in a state. That. The condition input processing unit 9 receives an input from the client 3 and passes it to the text mining processing unit 8 and provides a GUI to the client 3. The result output processing unit 10 includes a display and a display control unit that performs display control for displaying on the display. In practice, the processing unit generally performs control processing by developing a program stored in a memory on a RAM using a microcomputer such as a CPU.

  When the document database 5 is updated by the document database update processing unit 4, the document citation relation index table 6 is also synchronized. As a result, the document citation information included in the document is acquired and the document citation relation index table 6 is updated at the same time.

  FIG. 2 is a diagram illustrating a configuration example of a table stored in the document database 5. In the example illustrated in FIG. 2, a document ID 21, a title 22, a body 23, and a posting destination magazine name 24 for each document ID are stored in pairs. In this example, the title (title), text, and journal name are stored in the database as document information, but in addition, the abstract, author, year of publication, etc. may be added and stored instead of the text. It is also possible to store only the abstract.

  FIG. 3 is a diagram showing a configuration example of a document citation relation index table created by extracting only document citation information from the text information in the document database 5 according to the present embodiment. In the example illustrated in FIG. 3, the document ID 31 includes a set of a cited document ID 32 that is an ID of a cited document cited in each document ID 31. From this figure, for example, it can be seen that Document 1 is cited in Documents 11 to 14. It can also be seen that Document 12 is cited in Documents 21 and 21. In FIG. 3, only the relationship between directly cited documents is created in advance. When analyzing the related information of the cited document (document), it is possible to search the cited document ID field 32 for the original document ID.

  FIG. 4 is a diagram illustrating an example of a user interface for prompting input of keywords necessary for collecting documents. In the example shown in FIG. 4, the keyword input screen 41 includes a keyword input field group 42 necessary for verifying the experimental hypothesis and a button 43 for transmitting the input information to the server. Further, in the example shown in FIG. 4, in order to facilitate the input of keywords necessary for the verification of the experimental hypothesis, the user is prompted to input the keywords for each category such as field, experiment method, and event. Yes. In addition, you may add an experiment execution organization, the name of a contribution destination magazine, an author name, etc. as an input candidate. In addition, with respect to the keywords input for each category, for example, preferential treatment (high weighting) is added according to the degree of prominence for a famous posting destination magazine name, a prominent scholar, or the like. An interface may be prepared. After the keyword is input in the input field 42, the input keyword is sent to the server 1 by pressing the send button 43.

  FIG. 5 is a diagram illustrating an example of a screen that displays a text mining result and cited document information. As shown in FIG. 5, the result display screen 51 includes a document group 53 extracted and displayed in a list by text mining based on the input keyword, and the input first keyword (relationship between the first keyword and the first keyword). A certain second keyword may be included.) A feature of each document can be expressed by a two-dimensional matrix X1 in which the group 52 is displayed in a matrix form. In the display example shown in FIG. 5, the appearance frequencies of the first and second keywords are represented as numerical values for each document. Further, each document information for each document is displayed on the left side of the matrix X1. In each document information column, a field 54 indicating a document ID for identifying the document and a journal name of a posting destination, and a field 55 indicating an analysis result of the document citation relationship are displayed. Here, in the field 55, the number of cited documents is described at the end of the document 1, etc., and is the number of other documents cited in the document 1. For example, the number of documents cited in Document 1 is 3. In addition, for each cited document, the number of places cited in document 1 may be displayed on the next layer.

  The number of cited references indicates the number of documents when directly cited by other documents. For example, the number of other documents citing document 1 is 35, and it can be estimated that document 1 is a fairly important document.

  Based on the document information displayed on the screen 51 and its features, a button (View) 56 for performing a more detailed document search, and the user determines that the search result is necessary for each document ID. There is provided a button 59 for calling up a screen for displaying the relationship between the field 57 that can be checked and the document ID determined to be necessary by tracing the document citation relationship on the citation relationship. The display screen can be scrolled in the XY directions by using the scroll buttons 50a and 50b.

  The two-dimensional matrix X1 is configured to display the number of documents extracted by text mining for each keyword, so that the accuracy of the text mining processing result when attention is focused on a certain keyword is high when the number of extracted documents (keyword appearance frequency) is high. It can be assumed that the accuracy is high. Further, it is possible to analyze in which area of the document the result appears and display the result. For example, if it is known whether there are many occurrences in the explanation part of the prior art or whether the appearance frequency is high in an important part such as an experimental result, the estimation result regarding accuracy becomes accurate.

  Further, although not shown in the figure, for example, the total number of documents extracted by four keywords may be displayed so that a guideline of accuracy of extraction results by a certain keyword group can be obtained. Here, it is assumed that the total number of extracted documents is 100, 90, 70, 50 in order from Documents 1 to 4 (when the scroll button 50a is used to scroll right, this total number is displayed).

  In the display example of FIG. 5, the list is displayed in the order of documents including a large number of keywords input by the user (in this case, the total number) according to the result of the text mining search. In order to search for documents to be collected by analyzing existing documents, it may be possible to sort and display in the cited document field (55). For example, an area for setting the number of cited references is provided (may be provided in FIG. 4), and a sorting function for displaying only documents having a number of cited references of 30 or more (here, documents 1 and 3) is provided. Also good.

  6 to 8 are collected while evaluating the necessary documents by tracing the document citation relationship using the documents obtained as a result of the text mining search performed on the document database 5 using the keywords necessary for document collection. It is a figure which shows the display screen interface which shows a process, and demonstrates the flow of a process and a screen interface, referring this.

  FIG. 6 is an example of a screen that displays the result of text mining search using a specific keyword. In the display screen 51 shown in FIG. 6, as a result of the text mining search, information on documents citing document 1 having a document ID “1” that includes more keywords input by the user is displayed. For the purpose of reference, this is a display screen showing a state where a check is made in the check column 57 of document 1, this line is selected, and a button 56 (View) for performing a more detailed document search is clicked with the pointer 61. . The cited document ID display area 58 and the selected document related display button 59 will be described later.

  FIG. 7 is a transition screen from the screen of FIG. 6, and is a diagram showing an example of extracting and displaying a document group quoting the document ID “1” selected in FIG. is there. The document ID of the cited document is displayed in the field 71. In the list 72 indicating the document group quoting the document (ID1), the document 1 is displayed on the screen 51 displaying the search result. A list of 35 references cited is displayed (in the figure, only 4 references are shown, and all references can be viewed by scrolling). It is known that Document 11 and Document 12 obtain a significant number of citations (for example, 15 or more are significant). Here, the document 11 and the document 12 are selected by checking them in the check box 57, and a screen for grasping the situation in which the documents are collected by tracing the document citation relationship so far is displayed. The selected document related display button 59 is clicked by the pointer 73. It is also possible to check documents by paying attention to the number of documents extracted by a specific keyword. Note that the documents counted in the number of cited references are only directly cited documents.

  The meaning of the act of checking the documents 11 and 12 in FIG. 7 is the work of sequentially displaying the cited documents of a specific document as a list as shown in FIGS. By checking the documents that are candidates for the documents that you want to collect, you can sort out what citation relationship the selected documents were based on later.

  FIG. 8 shows a hierarchical display screen 81 of citation relationships that allows a bird's-eye view of the state of collecting documents by tracing the document citation relationships, and by specifying a specific document in the display screen 81, It is a figure which shows a mode that the display screen 86 which can display detailed information is displayed simultaneously. A hierarchy display screen 81 for citation relations for comprehensively grasping the state of collecting documents by tracing the citation relations of documents is a box 82 indicating a document composed of a document ID, a journal name to be posted, and the number of cited articles. And an arrow line 83 indicating a document citation relationship.

  In the document box 82 of this screen example, “Document 1” checked on the first list of text mining results (FIG. 6) is shown. All the relationships are displayed in FIG. Among them, the color of the document box is changed and displayed so that it is possible to organize and grasp which document was checked and selected. In the figure, the colors are changed in Reference 1, References 11 and 12, and Reference 22.

  Thus, a document box group 84 indicating a document group that indicates a document that is a source and that cites the document, and a document box 85 that indicates a document group that further cites the document is displayed on the screen 81. ing. An arrow indicating a cited relationship regarding a document can be created based on the document citation relationship index table shown in FIG. By preparing the document citation relation index table shown in FIG. 3 in advance, the display shown in FIG. 8 can be performed quickly. In addition, there is an advantage that processing for switching documents or switching display can be performed quickly. Here, regarding the document 12, as shown in the column 85, directly cited documents are associated by arrows. On this screen 81, the direct citation relationship between documents can be seen from a bird's-eye view by following the arrows.

  The display screen 86 is displayed by selecting / determining the document 12 with a pointer or the like on the display screen 81, for example. On the display screen 81, the same display as that of FIG. On this display screen 51, documents 21 to 24 (see FIG. 3) citing document 12 indicate the document 12 with arrows. In this way, in FIG. 8, it is possible to view the panoramic display screen 81 regarding the citation relationship and the detailed document information display 86 showing the citation relationship regarding a certain document and the result of the keyword search simultaneously or while switching.

  FIG. 9 is a flowchart showing the flow of processing in the text mining apparatus according to this embodiment.

  As shown in FIG. 9, in response to the user inputting a keyword for document collection from the client (step 91), the server receives the keyword transmitted from the client, and the input keyword is set to the mining condition. Then, the text mining process is executed on the document database (step 92).

  A document data list characterized by the input keyword is created by text mining, and after the cited document information (cited document and cited document) included in each document data is analyzed (step 93), the input keyword (and The keywords that are related to the input keyword) and the group of documents extracted by text mining show the characteristics of each document using a two-dimensional matrix, and display the results including the number of cited references and the number of cited references. (Step 94). It can be used as a reference for evaluation based on information on this screen. In addition, upon receiving an input from the user that they want to refer to the information of the cited reference, click on the detail display button of the cited reference on the screen to cite the article based on the specified cited reference. The relationship is analyzed, and the result including the number of cited references and the number of cited references of the document is displayed. When the necessary documents are collected by tracing the cited documents through these series of operations, the process of step 95 is repeated.

  Also, when this operation receives an input from the user that gives a bird's-eye view of the citation relationship of the documents referred to so far, the document information that has been referred to so far and their citation relationship are expressed in a bird's-eye view. The screen to be displayed can be displayed (step 96). A result display screen including the number of cited documents and the number of cited references of the document, and a screen that provides a bird's-eye view of the referenced document information and their citation relationship are displayed simultaneously or by switching. be able to.

  As described above, according to the text mining device according to the present embodiment, when collecting documents related to a specific theme, by searching for useful documents based on the number of cited references, together with searching documents by text mining. Useful literature can be collected. Further, by configuring the citation relationship so that it can be analyzed from a bird's-eye view, it is possible to save time and labor required for the document collection process.

  The present invention can be used for a text mining apparatus.

It is a conceptual diagram which shows the example of 1 structure of the text mining device by one embodiment of this invention. It is a figure which shows the example of 1 structure of the literature database by this Embodiment. It is a figure which shows the example of 1 structure of the literature quotation index information by this Embodiment. It is a figure which shows an example of the interface for prompting | inputting the keyword required for document collection in a text mining apparatus. It is a figure which shows an example of the screen which compoundly evaluates and displays the text mining result and the cited reference information of each document. It is a figure which shows the flow and screen interface of a screen interface for tracing and collecting required literatures by tracing a literature citation relationship. It is a display example which shows the information regarding the literature which cites the literature selected in FIG. 8 is a diagram showing a bird's-eye view of a citation relationship related to a selected document among the documents displayed in FIG. 7 and a display example showing information related to a document selected from the citation relationship. It is a flowchart figure which shows the flow of a process in the text mining device by this Embodiment.

Claims (4)

A text mining condition input processing unit that provides a user interface for prompting a client terminal to input a keyword for literature search, and that accepts a condition for performing a text mining search process based on the input keyword; ,
Based on the accepted conditions, a text mining processing unit that directly texts the document database,
Analyzing citation relation information of documents extracted based on text mining processing results based on a citation relation table created by extracting only citation information and arranging direct citation relations as pairs of documents. A document citation relationship analysis processing unit,
Based on the text mining process and the document citation relationship analysis process, the number of cited documents, the number of cited documents, and the number of documents extracted by text mining corresponding to a keyword are displayed separately for each document. A text mining device comprising: a display control unit that performs display control for each document. The display control unit
2. The text mining apparatus according to claim 1, wherein when a certain document is specified, hierarchical display control is performed to hierarchically display the citation relationship for a cited document having a relationship of quoting the document.   The text mining device according to claim 2, wherein the display control unit performs control to display the hierarchical display and the display by document on the same screen or by switching. Providing a user interface for prompting a client terminal to input a keyword for literature search, and receiving a condition for performing a text mining search process based on the input keyword;
Text mining directly into the bibliographic database based on the accepted conditions;
Creating a document citation relationship table created by extracting only the document citation information from the document citation information extracted by the text mining processing result;
Based on the text mining process and the document citation relationship analysis process, the number of cited documents, the number of cited documents, and the number of documents extracted by text mining corresponding to a keyword are displayed separately for each document. And a text mining method.