JP5446577B2 - Document analysis program and document analysis system - Google Patents

Description

  The present invention relates to a document analysis program and a document analysis system.

  Documents written in free form included in questionnaires and complaint documents contain a lot of useful information, so it is desirable to analyze the documents and feed them back to product development. However, because of the large amount, it is difficult to perform the analysis manually and in terms of time and cost, and attempts have been made to perform analysis by a computer.

  Patent Document 1 discloses creating a semantic network from word co-occurrence frequencies in a plurality of documents. Patent Document 2 discloses that a graph is created from a document, importance is assigned to a node of each graph, and relevance is assigned to a link, and these are compared between the graphs. Patent Document 3 discloses creating a graph and creating a subset based on the appearance frequency of words in a plurality of documents.

JP 2001-243223 A JP 2003-330966 A JP 2009-128949 A

  It is an object of the present invention to provide a means for comparing between different graphs targeting different document sets.

  The invention according to claim 1 is a graph data including a plurality of pieces of link data having character string information of words appearing in a document and node data having appearance frequency information about the words, and information linking a plurality of the node data. Is a document analysis program for causing a processing apparatus to execute a graph data selection procedure for selecting a plurality of graphs and a graph comparison display procedure for comparing and displaying the selected plurality of graph data.

  The invention according to claim 2 is a document input procedure in which the plurality of documents are input, and a plurality of words that have a high appearance frequency among the words included in the plurality of documents input in the document input procedure. A word extraction procedure for extracting a plurality of second words that are words having a high appearance frequency in a subset of documents including a first specific word that is one of the first words. The document analysis program according to claim 1, further comprising: a graph data storage procedure for storing link data having information that links the first specific word and the second word of the subset.

  According to a third aspect of the present invention, in the graph data selection procedure, a comparison item to be used in the comparison is further selected, and in the graph comparison display procedure, the comparison display is performed based on the selected comparison item. The document analysis program according to claim 1, wherein the document analysis program is a document analysis program.

  According to a fourth aspect of the present invention, in the graph comparison display procedure, for one graph data of the selected plurality of graph data, a node display including the character string information based on the node data The document analysis program according to any one of claims 1 to 3, wherein an arrow connecting nodes is displayed based on the link data.

  In the graph comparison display procedure, the character string information included in the node of the one graph data and the character string information in other graph data among the plurality of graph data may be included in the graph comparison display procedure. 5. The document analysis program according to claim 4, wherein the node or the character string information related to the matched character string information is displayed in an emphasized manner.

  According to a sixth aspect of the present invention, in the graph comparison display procedure, the appearance of the word relating to the same character string information that appears together in the one graph data and the other graph data among the plurality of graph data 6. The document according to claim 4, wherein the document is displayed with emphasis so as to indicate that the ratio of the appearance frequency in the other graph data is high or the ratio of the appearance frequency is low. It is an analysis program.

  According to a seventh aspect of the present invention, in the graph comparison and display procedure, the appearance of the word related to the same character string information appearing together in the one graph data and the other graph data among the plurality of graph data The frequency is displayed with emphasis so as to indicate that the rank of the appearance frequency is high or the rank of the appearance frequency is low in the one graph data. The document analysis program described in the section.

  According to an eighth aspect of the present invention, in the graph comparison and display procedure, the combination of the character string information combined by the information linking the node data included in one link data is the other of the plurality of graph data. If the combination of the character string information of the graph data is only matched with only one of the combinations of the character string information in any of the combinations of the character string information of the one graph data, The document analysis program according to any one of claims 4 to 7, wherein a node related to the other character string information of the combination of character string information or the character string information of the node is highlighted and displayed. is there.

  The invention according to claim 9 is a combination of two different character string information included in the one graph data and a combination of the character string information included in the other graph data in the graph comparison display procedure. If only one of the combinations of the character string information is a combination of the character string information that is combined by the information that links the node data included in the link data, Or the combination of the character string information is a combination of the character string information combined by the information that connects the node data included in the link data, but the combination order is different from each other The document analysis program according to any one of claims 4 to 8, wherein the arrow is highlighted and displayed.

  According to a tenth aspect of the present invention, the comparison result is displayed by superimposing and displaying other graph data of the plurality of graph data on the one graph data. The document analysis program according to claim 4, wherein the program is a document analysis program.

  The invention according to claim 11 is characterized in that when the superimposed display is performed, the node and the arrow are dynamically displayed by any of appearance, disappearance, discoloration, and enlargement / reduction. The document analysis program according to Item 10.

  The invention according to claim 12 is characterized in that the one graph data of the plurality of graph data includes other graph data to be compared. The document analysis program described in the section.

  According to a thirteenth aspect of the present invention, the display for emphasis is performed by changing any of color, line thickness, and line type, or by displaying in shadow, translucent, or bold. The document analysis program according to any one of claims 5 to 12, wherein the program is a document analysis program.

  The invention according to claim 14 is a graph data including a plurality of pieces of link data which is character string information of words appearing in a document, node data having appearance frequency information about the words, and information linking a plurality of the node data. To a graph selection unit that selects the plurality of graph data, and a comparison display control unit that compares and displays the selected plurality of graph data.

  The invention according to claim 15 is a plurality of words that have a high frequency of appearance among a document input unit to which the plurality of documents are input and each word included in the plurality of documents input by the document input unit. A word extraction unit that extracts a plurality of second words that are words having a high appearance frequency in a subset of documents including the first specific word that is one of the first words. The document analysis system according to claim 14, further comprising: a graph data storage unit that stores, in the graph data storage unit, information that links the first specific word and the second word of the subset. It is.

  According to the invention described in claims 1 and 14, a plurality of different graph data targeting a set of different documents can be compared and displayed.

  According to the second and fifteenth aspects of the present invention, it is possible to realize a series of processes from inputting a plurality of documents to comparing and displaying graphs.

  According to the third aspect of the present invention, it is possible to comparatively display comparison items as necessary.

  According to the fourth aspect of the present invention, the comparison display can be performed based on the display of the node and the arrow for one graph data.

  According to the invention described in claim 5, it is possible to compare and display nodes having matching character string information among a plurality of graph data.

  According to the sixth aspect of the present invention, it is possible to compare and display the difference in the ratio of the appearance frequencies of the same word that appears together in one graph data and the other graph data.

  According to the seventh aspect of the present invention, it is possible to compare and display the difference in the ranks of the appearance frequencies of the same word that appears together in one graph data and the other graph data.

  According to the eighth aspect of the present invention, it is possible to compare and display words having different co-occurrence partner words among a plurality of graph data.

  According to the ninth aspect of the present invention, it is possible to compare and display that the relationship between the co-occurring partner words is different among a plurality of graph data.

  According to invention of Claim 10, 11 and 13, a difference can be displayed more visually conspicuously.

  According to the twelfth aspect of the present invention, it is possible to visually display one piece of graph data and a portion of the graph data included therein.

It is a figure which shows the document analysis system which concerns on one Embodiment of this invention. It is a figure which shows the functional block of the document analysis system of FIG. It is a flowchart which shows a graph data creation process. It is a figure which shows the data structure of word appearance data. It is a figure which shows the data structure of node data. It is a figure which shows the data structure of link data. It is a flowchart which shows a graph data comparison display process. It is a figure which shows a part of graph data selection screen. It is a figure which shows a part of comparison item selection screen. It is a graph of single graph data P. It is the graph highlighted about the same word of graph data P and Q. It is the graph highlighted about the ratio difference of the appearance frequency of graph data P and Q. It is the graph highlighted about the order of appearance frequency of graph data P and Q. It is the graph highlighted about the difference of the co-occurrence partner of graph data P and Q. It is the graph highlighted about the relationship between the co-occurrence partners of the graph data P and Q. The graph is highlighted for graph data Q when graph data P is a set of all graph data.

  FIG. 1 shows a document analysis system 100 according to an embodiment of the present invention. The document analysis system 100 includes a computer main body 200 including a CPU (Central Processing Unit), a RAM (Random Access Memory), a magnetic disk device, and the like, a display device 102 that performs screen display according to a command from the computer main body 200, and a computer main body 200. And an input device 104 for inputting information. Here, the display device 102 may be any of a liquid crystal display device, a CRT (Cathode Ray Tube) and other display devices, and the input device 104 includes a key input device, a pointing device such as a mouse, and a scanner. An image input device is included.

  FIG. 2 shows a functional block diagram of the document analysis system 100. Here, the computer main body 200 functions by executing a program stored in an internal magnetic disk device, and the functional blocks 202 to 210 in FIG. 2 are realized by the program. As shown in FIG. 2, the computer main body 200 analyzes a document input unit 206 that receives data of a document input from the input device, a document input in the document input unit 206, and a word included in the document and its word A word extraction unit 210 that extracts both appearance frequencies, and a graph data creation and storage unit that creates node data 254 and link data 256 (to be described later) for the words extracted by the word extraction unit 210 and stores them in the graph data storage unit 212 218, a graph selection unit that displays graph data for comparison and a screen for selecting comparison items on the display device 102, and obtains graph data selected by input from the input device 104 from the graph data storage unit 212 202 and the graph data selected by the graph selection unit 202 and the contents of the comparison items are acquired and displayed. And a comparison display controller 204 to display in comparison with the location 102.

  FIG. 3 shows a flowchart of the graph data creation process. In the graph data creation process, first, a document is input in the document input process in step S101. A document is input by inputting text information to the document input unit 206 via the input device 104. In addition to a key input device as the input device 104, a scanner, etc., a computer device connected to a network, etc. A document may be input. In the present embodiment, a questionnaire result regarding a mobile phone is input as an input document.

  Next, in the word extraction process in step S102, the word extraction unit 210 extracts words. In this word extraction process, word appearance data 252 having the data structure shown in FIG. 4 is created. In the word appearance data 252, as shown in FIG. 4, a message number assigned to a document, a word ID that is an identifier of the word, and a message attribute such as the age, sex, and region of the message creator are recorded. Attribute 1, attribute 2, and attribute 3.

  In the graph data creation / save process of step S103, the graph data creation / save unit 218 identifies words with high appearance frequency from the word appearance data 252 in a set of a plurality of documents (referred to as “whole set”). Furthermore, for a set of documents including the specified high-frequency words (referred to as “subset”), it is repeated to specify other high-frequency words, that is, words that co-occur with high frequency, Node data 254 as shown in FIG. 5 is created for each identified word. The node data 254 includes a node ID, a word ID, a word, and the number of elements, as shown in the data configuration of FIG.

  Further, as shown in the data configuration of the link data 256 in FIG. 6, the graph data creation / storing unit 218 further sets the node ID of the identified high-frequency word as the source node ID, and identifies the identified high occurrence. The link data 256 is generated in which a link ID is added to a combination of the node ID of the other high-frequency words identified in the subset including the frequency word, that is, the node ID of the co-occurrence partner word as the target node ID. The created node data 254 and link data 256 are saved as graph data in the graph data storage unit 212 by the graph data creation / save unit 218.

  Next, the graph data comparison display process S200 will be described. FIG. 7 shows a flowchart of the graph data comparison display process S200. The graph data comparison display process S200 includes a selection process (step S210) and a comparison display process (step S220). The selection process in step S210 further includes a graph selection process in step S212 and a comparison item selection in step S214. Processing.

  First, in the selection process in step S212, the graph selection unit 202 causes the display device 102 to display a screen that allows the user to select a plurality of graph data to be compared. FIG. 8 shows a part of the selection screen. When the graph data is selected by the user, a plurality of selected graph data is acquired from the graph data storage unit 212. In the present embodiment, it is assumed that the graph data P and the graph data Q are selected, but the number of objects for comparison display may be three or more. Here, the graph data P and the graph data Q are a set of questionnaire results regarding mobile phones performed in different regions.

  Next, in the comparison item selection processing in step S214, the graph selection unit 202 displays the type of comparison item on the display device 102 in order to cause the user to select the comparison item highlighted in the graph data to be compared. Let FIG. 9 shows a part of the selection screen. The types of comparison items include the same word, the difference in the appearance frequency of the same word, the rank of the appearance frequency of the same word, the difference in co-occurrence partners, the difference in co-occurrence relationship, and the like. Here, the “co-occurrence combination” is selected. Subsequently, in the comparison display process in step S220, the comparison result is displayed based on the graph data and the comparison items selected in the selection process in step S210.

  FIG. 10 shows a case where the graph of the graph data P is displayed alone. In this graph, the words “phone”, “mobile phone”, “PHS”, “necessary”, and “mail”, which are words having a high appearance frequency, are shown in the document included in the entire set A of the graph data P. Of these, the words “convenience”, “manner” and “phone”, which are words having a high appearance frequency in a subset of documents including the word “mobile phone”, are shown.

  FIG. 11 shows a case where the “same word” selected as the comparison item is highlighted in the comparison display that highlights the difference from the graph data Q while displaying the graph of the graph data P. . As shown in this figure, among the words included in the nodes of the graph data P, the words “phone”, “mobile phone”, “PHS” and “ “Useful” is highlighted by thickening the line indicating the frame of the node.

  FIG. 12 shows a comparative display emphasized when the difference in the frequency of appearance of the same word is different. The word of the node whose frame is drawn with a bold line has a higher ratio of the appearance frequency in the graph data Q than the ratio of the appearance frequency in the graph data P, and the word of the node whose frame is drawn with a dotted line is the graph data Q It means that the ratio of the appearance frequency in is lower than the ratio of the appearance frequency in the graph data P. In other words, the word “phone” is more frequent in the graph data Q than in the graph data P, and the words “mobile phone”, “convenience”, and “PHS” are in the graph data Q more than the graph data P. The frequency of appearance is low.

  FIG. 13 shows a comparative display that is emphasized when the appearance frequency ranks of the same word are different. In the graph of FIG. 12, in the graph data Q, the word “PHS” has a higher appearance frequency than the word “telephone”, but in the graph data P, the opposite is true for the word “PHS” and the word “phone”. "Is shown by adding a shadow to the node and changing the color of the shadow. For example, in the graph data P, when the third highest occurrence frequency is the word “phone” and the fifth highest is the word “PHS”, in the graph data Q, the word “phone” is the fifth highest appearance frequency, This is the case where “PHS” is third.

  FIG. 14 shows a case where there is a difference in co-occurrence partners. In the graph of FIG. 13, when the word “charge” as the co-occurrence partner of the word “mobile phone” exists in the graph data Q, but does not exist in the data P, both the arrow and the node frame are indicated by dotted lines. By showing.

  FIG. 15 shows the case where the direction of the arrow, that is, the relationship between the source node and the target node is different for words having the same combination of co-occurrence. In this graph, only the arrow from the word “mobile phone” to the word “convenient” exists in the graph data P, but the graph data Q includes not only the arrow from the word “mobile phone” to the word “convenience”, but also the word In the case where an arrow from “convenient” to the word “mobile phone” is also present, the tip of the arrow from the word “convenient” to the word “mobile phone” is greatly emphasized. In addition, it is possible to perform highlighting when there is no arrow of one graph data or when the direction of the arrow is reversed.

  FIG. 16 shows a case where a set of all graph data is set as a comparison target. In the figure, nodes of all graph data V (set G) including the graph data Q to U are shown, and nodes and arrows of the graph data Q to be compared with all the graph data are indicated by bold lines, and nodes The font in is changed from the font in other nodes.

  In the above-described embodiment, highlighting is limited to bold lines, dotted lines, shadows, and bold characters due to drawing restrictions and the like. However, in the screen display, highlighting by changing colors, semi-transparent display, and nodes and arrows Can be highlighted using dynamic changes such as the appearance, disappearance, discoloration and scaling. In addition, it is possible to make it easier to recognize the difference due to the presence or absence of the comparison display by installing a button on the screen whether or not to perform the comparison display and performing a highlight display when a command to perform the comparison display is issued. Good.

  Moreover, in the above-mentioned embodiment, although the text of the questionnaire result of the mobile phone is taken as an example, the object of analysis is not limited to this, and it may be a set of other texts, and can already be displayed in a graph. It may be data about a certain word.

  In the above-described embodiment, the system is configured by one apparatus. However, each component may be stored in an apparatus via a network to configure one system.

  In the above-described embodiment, the program is stored in the magnetic disk device. However, the program may be provided by being stored in another storage medium such as a CD-ROM.

  DESCRIPTION OF SYMBOLS 100 Document analysis system, 102 Display apparatus, 104 input apparatus, 200 Computer main body, 202 Graph selection part, 204 Comparison display control part, 206 Document input part, 210 Word extraction part, 212 Graph data storage part, 218 Graph data creation preservation | save part 252 word appearance data, 254 node data, 256 link data.

Claims (13)

Document input procedure for entering multiple documents,
Among each word included in the whole set that is the whole of the plurality of documents input in the document input procedure, a plurality of first words that are words having a high appearance frequency are extracted, and the whole of the plurality of documents is extracted. A word extraction procedure for extracting a plurality of second words that are words having a high appearance frequency in a subset that is a set of documents including the first specific word that is one of the first words;
A graph data storage procedure for storing link data having information for connecting the first specific word and the second word of the subset;
Graph data selection procedure for selecting a plurality of graph data each including a plurality of link data having node data having character string information of words appearing in a document and appearance frequency information for the words, and information linking a plurality of the node data When,
A document analysis program for causing a processing device to execute a graph comparison display procedure for comparing and displaying the selected plurality of graph data. In the graph data selection procedure, further select a comparison item to be used in the comparison,
The document analysis program according to claim 1, wherein the comparison display is performed based on the selected comparison item in the graph comparison display procedure.   In the graph comparison display procedure, for one graph data of the selected plurality of graph data, a node display including the character string information based on the node data and an inter-node based on the link data The document analysis program according to any one of claims 1 to 2, wherein an arrow connecting the two is displayed.   In the graph comparison display procedure, the character string information included in the node of the one graph data and the character string information in the other graph data among the plurality of graph data are related to the character string information that matches. The document analysis program according to claim 3, wherein the node or the character string information is highlighted and displayed.   In the graph comparison display procedure, in the appearance frequency of the word related to the same character string information appearing together in the one graph data and the other graph data among the plurality of graph data, in the other graph data The document analysis program according to claim 3 or 4, wherein the document analysis program is displayed with emphasis so as to indicate that the appearance frequency ratio is high or the appearance frequency ratio is low.   In the graph comparison display procedure, in the appearance frequency of the word related to the same character string information appearing together in the one graph data and the other graph data among the plurality of graph data, in the one graph data The document analysis program according to any one of claims 3 to 5, wherein the document analysis program is displayed with emphasis so as to indicate that the rank of the appearance frequency is high or that the rank of the appearance frequency is low.   In the graph comparison display procedure, for the combination of the character string information combined by the information linking the node data included in one link data, the character string information of the other graph data of the plurality of graph data If the combination matches only one of the combinations of the character string information in any combination of the character string information of the one graph data, the other character of the combination of the character string information The document analysis program according to any one of claims 3 to 6, wherein a node related to column information or character string information of the node is highlighted.   In the graph comparison display procedure, the combination of two different character string information included in the one graph data and the same character string information combination as the combination of the character string information included in the other graph data Among the combinations of the character string information, only one of the combinations of the character string information is a combination of the character string information combined by the information connecting the node data included in the link data, or the combination of the character string information Both are combinations of the character string information combined by the information connecting the node data included in the link data. When the combination order is different from each other, the arrow is highlighted. The document analysis program according to any one of claims 3 to 7, wherein   The display of the compared result is performed by superimposing and displaying other graph data of the plurality of graph data on the one graph data. The document analysis program according to any one of the above.   10. The document analysis program according to claim 9, wherein, when the superimposed display is performed, the node and the arrow are dynamically displayed by any of appearance, disappearance, discoloration, and enlargement / reduction.   The document analysis program according to any one of claims 3 to 10, wherein the one graph data of the plurality of graph data includes other graph data to be compared.   The display for emphasis is performed by changing any of color, line thickness, and line type, or by displaying in shadow, translucent, or bold. The document analysis program according to any one of the above. A document input unit multiple documents are inputted,
Extracting a plurality of first words, which are words having a high appearance frequency, from among the words included in the whole set that is the whole of the plurality of documents input by the document input unit, and out of the whole of the plurality of documents And a word extraction unit that extracts a plurality of second words that are words having a high appearance frequency in a subset that is a set of documents including the first specific word that is one of the first words;
A graph data storage unit for storing link data having information for connecting the first specific word and the second word of the subset;
Node data, and a plurality of said nodes graph data selecting unit data the link data with information linking multiple select multiple including Gurafude data each having a frequency of occurrence information for the word string information and the words appearing in the document And a comparison display control unit for comparing and displaying the selected plurality of graph data.

Images