JP4477587B2 - Method for generating operation buttons for computer processing of text data - Google Patents

Description

  The present invention recognizes the meaning of text data and automatically generates operation buttons (meaning buttons) for searching the text data in a multidimensional manner and generates operation buttons for computer processing of the text data to be displayed on the display device. Regarding the method.

  In the field of text data processing by computer, mainly in the field of text mining and knowledge management, keyword extraction, semantic recognition by general-purpose dictionary, keyword categorization, and these results were utilized by the development of Japanese language processing technology. Various types of analysis and utilization are starting to take place. In this analysis, for example, the part-of-speech appearance frequency, keyword appearance frequency, the relationship between keywords and data attributes (when, who, where, etc.), the relationship between keywords, the time series change of keyword appearance frequency, etc. are output in various formats. It is now possible to use these outputs to make use of them by application.

  However, in the conventional technology, (a) the result of the analysis depends on the accuracy of the dictionary, (b) does not fit a specific use or purpose, (c) has a function of updating the dictionary, There is a problem that it is difficult to respond to each business need. For example, it is usually included in text data such as recognition of the company's product name, employee name, department name, distinction between the company's product and other company's product, recognition of the name of the product function or specification, recognition of the company name or name of the business partner, etc. The user-specific keywords cannot be handled with a general-purpose dictionary. Even if the user can add these keywords to the dictionary, the keywords can be defined according to multiple tasks and usages, and the meaning recognition criteria can be defined and used freely. It is not. For example, even with the same keyword in notation, the meaning and interpretation are often different depending on the field used, but such a flexible recognition cannot be performed with a general-purpose dictionary.

  In addition, the conventional technique using a general-purpose dictionary is difficult to cope with. For example, in text data that collects customer inquiries by telephone, fax, e-mail, or oral, customer inquiries (text data) are classified into “complaints”, “questions”, “requests”, etc. Furthermore, there is a need to further subdivide each and perform various analyses. In this business, what kind of words and expressions are basically judged as “complaints”, how much words and expressions that are synonymous or similar to the above judgments should be included in “complaints”, However, it is necessary to determine the judgment criteria that are suitable for the purpose of the business, such as what words and phrases should be excluded. These needs are not something that can be determined by pulling a traditional dictionary.

In general fixed-length record format data in databases and files, the operation button is automatically selected so that the end user can easily search, analyze and utilize the desired data simply by selecting the operation button. The technology to generate has been developed (see Patent Document 1). However, this technology cannot handle text-format data that is a collection of variable-length sentences, and the end user simply selects an operation button and wants it based on the meaning of the selected operation button. It is not a technology that allows text data to be easily searched, analyzed, and utilized.
Japanese Patent No. 2702416

  Information processing that makes it easy for end users to analyze text data without specialized knowledge, that is, operation buttons based on the meaning and attributes of text data are automatically displayed and only those buttons are selected. Therefore, there is a need for information processing that allows text data to be narrowed down in a multi-dimensional hierarchy, or displayed in association with parallel data, so that it can be freely searched, analyzed, and utilized.

  For such text data analysis, a new mechanism is required to replace the conventional method of meaning recognition by general-purpose dictionary lookup, which is a universal knowledge base. In other words, by utilizing know-how based on user-specific knowledge, experience, and wisdom, text data that can be easily and individually responded to changes in the environment and circumstances, and can be applied to both general and special purposes is analyzed. In other words, a mechanism for processing based on a concept such as a “semantic recognition rule” is required.

  Semantic recognition is a mechanism that makes it possible to analyze the form of text data based on a rule (semantic recognition rule) and distribute the text data into categories defined by the rule. For example, if text data is assigned to the category “complaint” and category “printer” defined in the rule, the text data is recognized as having a meaning related to “complaint” and “printer”. It is a mechanism that thinks.

  Specifically, first, the category is defined by associating a specific character string (specific keyword) representing a meaning unique to the user with an arbitrary category according to individual uses of a plurality of businesses and industries. There is a need for a mechanism that allows a set of categories to be freely and easily configured into arbitrary rules. Specific keywords include, for example, the company's product name / employee name / department name, other company's product name, important function name / specification name of the product, and the company name / name of the supplier.

Similarly, a category is defined by associating a common language element (conceptual keyword) included in a phrase or expression with a meaning that is important to the user and associated with an arbitrary category according to individual usages of a plurality of businesses and industries. However, there is a need for a mechanism that allows a set of such categories to be freely and easily configured into arbitrary semantic recognition rules. The conceptual keywords, for example, "not start", "can not be displayed", "unless return", "no ~" is the language elements contained in common expressions such as "not be printed", or its There are cases where there is a utilization form and it is desired to analyze in association with the category of “complaints” in this business.

  Furthermore, based on an arbitrary rule (semantic recognition rule) configured by integrating two mechanisms of a specific keyword and a conceptual keyword, the form of text data is analyzed, the two types of keywords are extracted, and the extracted keywords It is necessary to provide a mechanism for automatically generating operation buttons in which the meanings are assigned to the corresponding categories and the categories and keywords, the keywords and keywords of other categories, and the keywords and text data are associated with each other.

  In the invention described in the cited document 1 relating to the operation button, it is possible to automatically generate the operation button based on the data configured by the field corresponding to the category and analyze the data. However, in the case of text data, a keyword corresponding to the category is a keyword included in the text data, but this keyword is not determined in advance like a field. For this reason, when trying to associate with the corresponding category according to the meaning of the keyword included in the text data, it is not known in advance how many keywords the text data includes. Therefore, an unspecified number of categories are associated with the result obtained by recognizing the meaning of the text data, and cannot be handled by the invention described in Patent Document 1 based on data in which fields are determined.

  The present invention has been made in view of the above circumstances, and automatically generates operation buttons (meaning buttons) based on various meanings of text data and displays them on a display device. From various viewpoints corresponding to categories and keywords, you can search while dynamically narrowing down the hierarchy in multiple dimensions, and redisplay multiple button classes and individual buttons in parallel to correlate with keywords of other categories It is an object of the present invention to provide an operation button generation method for computer processing of text data that can be searched while viewing.

In order to achieve the above object, an operation button generation method for computer processing of text data according to the present invention retrieves arbitrary text data from a plurality of text data including a character string and having a file name and stored in a storage device . A computer data processing button generation method for text data generated by computer processing for a computer processing operation button for the purpose, and a representation element for matching a category and a character string in the text data A keyword is defined in association with each other, and text data in which the keyword is included in a character string in the text data is searched, and it is found that the category, the keyword, and the keyword are included . the text data or et al., category field, key Possess three fields of the word fields and text data file name field, each field, the category, corresponding the keyword, and the keyword file name of the text data that is retrieved that contain the respectively one-to-one Creates a button of source data having a meaning recognition result table is a set of result records obtained by the analytical button cLASS in correspondence with the field value of the category field of the button of the source data, the field of the keyword field the individual button belonging to the button class for the analysis generated each in correspondence to a value, the Browse button cLASS from the text data file name field of the button of the source data to correspond to a field value of the text data file name field The individual button belonging to the button class for the reference product, respectively, for displaying the button class and the individual buttons on the display device.

  Thus, operation buttons (meaning buttons) can be automatically generated and displayed on the display device based on various meanings of the text data from various viewpoints corresponding to arbitrary categories and keywords.

The analysis button class and the reference button class are displayed in parallel on the display device together with the individual buttons, and when any individual button belonging to one analysis button class is selected, the selected individual button corresponds to the selected individual button The result record including the keyword corresponding to the field value of the keyword field and at least one of the other keywords included in the text data including the keyword is extracted from the semantic recognition result table, and the extracted result record Preferably, the individual buttons belonging to the analysis button class and the reference button class are generated and displayed again.
Thus, each time an individual button is selected, the correlation between keywords (individual buttons) between other categories (button classes) can be known.

Meaning the result record including all the keywords respectively corresponding to the field values of the keyword field corresponding to the selected individual button by selecting any individual button of any button class for analysis in any order It is preferable to extract from the recognition result table, and generate and re-display the individual buttons belonging to the analysis button class and the reference button class based on the extracted result record .
As a result, text data having a specific meaning can be searched for or the meaning of the specific text data can be examined simply by allowing the user to select the displayed operation button (meaning button).

The semantic recognition result table defines, for example, the keywords and defines the categories while associating the categories and keywords with each other by defining the keywords and individually defining references to the defined keywords for each category. A keyword that matches the keyword is extracted from among the keywords, associated with a category in which a reference to the keyword is defined, and generated based on the associated result.

The keywords are categorized and defined as a specific keyword composed of a specific character string and a concept keyword composed of a character string including an abstracted portion. In the specific keyword, the specific character string is defined as the concept. In the keyword, it is preferable that a character string excluding the abstracted portion from a character string including the abstracted portion is used for matching with a character string in the text data .
A keyword is an expression element that matches a character string in text data. The keyword (expression element) is classified into a specific keyword that is a specific character string and a concept keyword that includes an abstracted character string. By defining, a specific character string (specific keyword) expressing a meaning unique to the user and a common language element (concept keyword) included in a phrase or expression that expresses an important meaning for the user are associated with each other. Operation buttons can be automatically generated.

  The keyword preferably includes a keyword name, an individual keyword, and an individual excluded keyword, and the individual keyword is searched as a synonym for the keyword name while excluding the individual excluded keyword.

  According to the present invention, it is possible to define as many keywords and categories as necessary corresponding to the purpose and purpose of business, so that operation buttons can be automatically generated based on various meanings of text data. . As a result, even end users with little computer experience and no specialized knowledge can select a large number of text data from various viewpoints corresponding to arbitrary categories and keywords. It is easy to search while narrowing down, and it is also possible to search while looking at the correlation with keywords of other categories by redisplaying multiple button classes and individual buttons in parallel. Can be analyzed and utilized.

  In the invention described in Patent Document 1, since button generation is performed using source data in which a field corresponding to a category can be specified in advance, when a field is unspecified, that is, what category appears in the source data. It cannot be applied when it cannot be determined in advance. According to the present invention, it is possible to automatically generate a button for source data in which an appearing category cannot be determined.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. As shown in FIG. 1, the computer system for carrying out the present invention, that is, the hardware of the text data analysis system using operation buttons shown below, includes a central processing unit 10, a storage unit 12, a display unit 14, and an input unit 16. It is mainly composed.

  As shown in FIG. 2, the software of the text data analysis system using the operation button determines the meaning of the text data according to the meaning recognition rule set by the meaning recognition rule definition tool 20 and the meaning recognition rule definition tool 20, and generates a button source. Semantic recognition tool 22 for creating data, from the recognition result obtained by the semantic recognition tool 22, the classification (category) of meaning, the element (keyword) and the text data that have been given meaning are associated with each other, and specialized knowledge is required. The meaning button automatic generation tool 24 for generating an operation button (meaning button) not to be displayed, and the operation button (meaning button) generated by the meaning button automatic generation tool 24 are displayed, and the user selects the displayed operation button. Just search for text data with a specific meaning, It consists of four tools of analysis tool 26 by means button to make it possible to examine or the meaning of the text data of.

  Hereinafter, the information processing apparatus and the program composed of the four means of the semantic recognition rule definition tool 20, the semantic recognition tool 22, the semantic button automatic generation tool 24, and the semantic button analysis tool 26 are collectively referred to as “text by operation button”. Called “data analysis system”.

  Here, the semantic recognition rule definition tool 20 provides an operating environment for registering a “meaning recognition rule”, which is a reference for giving meaning to text data, by the user using the input device 16 and interacting with the computer. To do. One semantic recognition rule is composed of each element of “keyword definition part”, “category definition part”, “applied category designation part”, and “rule name designation part”. It has a keyword definition function 20a, a category definition function 20b, an application category specification function 20c, and a rule name specification function 20d for executing definition or specification of elements, respectively. The user can register an arbitrary number of “meaning recognition rules” with the meaning recognition rule definition tool 20, save it as a “meaning recognition rule file” in the storage device 12, and update, reference, and reuse as necessary. .

  The “keyword definition section” is composed of an arbitrary number of “keywords”, and the keyword definition function 20a has a function for defining keywords. The keyword is an expression element for matching with a character string in the text data. In this example, there are two kinds of “concept keyword” including a specific character string “specific keyword” and an abstract character string. I know.

  A “category definition section” is defined by the category definition function 20b. An arbitrary number of “categories” constitute the category definition section. An arbitrary number of keywords are associated with one category by the category definition function 20b. The keyword associated with the category is a basis for sorting the text data including the keyword into the category by the semantic recognition tool 22.

The application category designation function 20c is a function that allows the user to designate which of the predefined categories is used for semantic recognition, and the designated category is stored in the “application category designation section”.
The “rule name designating part” is composed of an arbitrary character string and uniquely identifies a rule. The user designates a “rule name” using the rule name designation function 20d.

  The meaning recognition tool 22 uses the input device 16 and the user interacts with the computer to analyze the form of text data and extract keywords according to the contents of the meaning recognition rules, and to extract the text data based on the extracted keywords. Is provided by the category, and an operating environment for creating buttoned source data based on the result is provided. The meaning recognition tool 22 has an analysis target text data specification function 22a, a meaning recognition rule specification function 22b, and a buttoned source data creation function 22c.

  The user designates text data to be analyzed by the analysis target text data specifying function 22a. “Text data” is a set of “text units”. One text unit includes a body composed of a character string, and an arbitrary number of other attribute information as required. Further, the user selects and designates one rule to be applied to the analysis target text data from the semantic recognition rule file created by the semantic recognition rule definition tool 20 by means of the semantic recognition rule specification function 22b.

  The buttoned source data creation function 22c is a function for executing button recognition source data by executing the meaning recognition process using the specified analysis target text data and the meaning recognition rule, and is defined in the meaning recognition rule. A keyword is extracted from the text data to be analyzed, and the text unit is assigned to a category associated with the keyword. Based on the result, “buttonized source data” is created and stored in the storage device 12 as a “buttonized source data file”. As shown in FIG. 10, the buttoned source data includes a semantic recognition rule file name 52, an analysis target text data storage location 54, and a semantic recognition result table 56. The meaning recognition result table 56 is composed of a set of result records 58 having at least three fields of a category 58a, a keyword 58b, and a text data file name 58c.

  The meaning button automatic generation tool 24 provides a function of automatically generating operation buttons (meaning buttons) for enabling the end user to easily analyze the meaning of the text data based on the button source data. The operation buttons are composed of individual buttons corresponding to keywords and the like, and button classes corresponding to categories, text data, and the like. The meaning button automatic generation tool 24 has a button class generation function 24a and an individual button generation function 24b.

  The button class generation function 24a scans the button source data, extracts elements that can be button classes, and generates a button class. The button class has a button class name and manages individual buttons associated with the button class. Buttoned source data includes elements that can become two types of button classes. First, a button class can be a unique value appearing in a category field of the meaning recognition result table, that is, an individual category name is a button class name. Secondly, a button class can be a field excluding a category field and a keyword field in the meaning recognition result table, and this is called a “reference field”. As shown in FIG. 10, if the result record 58 is composed of, for example, three fields of a category 58a, a keyword 58b, and a text data file name 58c, the field of the text data file name 58c is the reference field. For example, the button class is named “target”. Hereinafter, the former is referred to as “analysis button class”, and the latter is referred to as “reference button class”. Therefore, the number of button classes is the sum of the number of analysis button classes and the number of reference button classes.

  The individual button generation function 24b generates an individual button by associating a unique value appearing in the field of the keyword 58b of the button source data 50 shown in FIG. And a unique value appearing in the reference field, that is, an individual data is generated as an individual button name and associated with the reference button class.

  In the invention described in Patent Document 1, a button is generated based on record-format data composed of a plurality of fields. According to this method, for one piece of data, an individual button belonging to the button class corresponding to the buttoned field is always generated, and one individual button is always generated corresponding to one field. . On the other hand, in this example, a button is generated based on text (sentence) format data including a keyword defined in the semantic recognition rule, and whether or not the keyword appears for one piece of data. The number of buttons when they appear, the number of individual buttons belonging to which button class are generated, and the number of buttons generated even when they are generated are completely undefined. In some cases, no individual button belonging to any button class is generated for a single piece of text data. In this way, in this example, the format of data on which buttons are generated is fundamentally different from existing patents, and button classes and individual buttons are generated in completely different ways based on the data in the different formats. .

  The meaning button analysis tool 26 displays the meaning button generated by the meaning button automatic generation tool 24 on the display device 14 and allows the user to select an arbitrary button using the input device 16. The semantic button analysis tool 26 has a semantic button parallel relation display function 26a and a text data search and content display function 26b.

The meaning recognition result table holds a relationship between a button class and individual buttons belonging to the button class, and a relationship between individual buttons that straddle the button class. The meaning button parallel relation display function 26a is a function for reflecting the relation included in the result table in the meaning button operation of the user, and when a certain individual button is selected, individual button classes related to the individual button are individually displayed. Since the buttons are redisplayed at the same time, information related to the selected individual button can be instantly confirmed and easily retrieved.
The text data search and content display function 26b is a function for searching text data of an arbitrary category or extracting a keyword included in the text data and displaying it by simply selecting a button by the user. It is possible to easily perform various analyzes such as grasping the correlation between categories and keywords. It is also possible to easily display the details of the text data narrowed down by selecting a button and the keywords included in the narrowed-down target.

Next, the text data analysis system using the operation buttons shown in FIG. 2 will be described in more detail with reference to FIGS.
The format of the text data to be analyzed used in this example is assumed as follows. The text unit is a single file stored in the storage device 12 of the computer. The content of this file is Japanese consisting of single or plural sentences, and a key for uniquely identifying the text unit is used. The file name. In addition to the above, the text data format may be a relational database table or various text resources on the Internet. These can also be analyzed in the same way as the text data format assumed in this example. is there. In this example, it is assumed that a text data analysis system using operation buttons is used at a user support window of a certain PC peripheral device manufacturer.

  FIG. 3 shows an example of text data used as an analysis target. As shown in the legend in FIG. 3, the text unit is composed of a file name and the contents of text data. Assume that these files (text data) having file names such as “text A” and “text B” are stored in a location “file: / text / user_support /” in the storage device 12 of the computer.

  FIG. 4 shows the structure of the semantic recognition rule 30 created by the semantic recognition rule definition tool 20 shown in FIG. In FIG. 4, “1” indicates one, “*” indicates one or more, and “**” indicates zero or more. One rule name 32, one keyword definition unit 34, one category definition unit 36, and one application category designation unit 38 correspond to one meaning recognition rule 30. In the following example, it is assumed that the rule name is designated as “user support analysis rule”.

  The keyword definition unit 34 includes one or more specific keywords 40 and one or more concept keywords 42. The specific keyword 40 includes one specific keyword name 40a, one or more specific individual keywords 40b, and zero or more specific individual exclusion keywords 40c. The concept keyword 42 includes one concept keyword name 42a, one or more concept individual keywords 42b, and zero or more concept individual exclusion keywords 42c. The category definition unit 36 includes one or more categories 44. In one category 44, one category name 44a and a reference 44b to one or more keywords (specific keywords or concept keywords) defined in the keyword definition section are arranged. The application category designating unit 38 includes references 46 to one or more categories already defined by the category defining unit.

  Table 1 is a definition example of the semantic recognition rule 30. To make it easier to understand, it is basically shown in XML format. Note that some of the closing tags are omitted to avoid complications. In Table 1, first, the rule name 32 is designated as “user support analysis rule”. Next, in the keyword definition part, <keyword k_id = “k-001” type = “concept” name = “not” ”is defined as the first keyword. k_id is an identifier for uniquely identifying an individual keyword, and “k-001” is designated here. “Type” is a type of keyword, and “concept” is designated here, and this keyword is a concept keyword. Furthermore, “name” is a keyword name, and “to not” is designated here. The concept keyword of “not” is followed by definitions of concept individual keywords and concept individual exclusion keywords. First, <individual term = "no" "/> representing a concept individual keyword is defined. The term is the definition content of the concept individual keyword, and “not” is designated here. Secondly, <individual term = “not” ”/> representing a concept individual keyword is defined. Third, <individual exclusion term = “no problem” /> representing a concept individual exclusion keyword is defined. term is the definition content of the concept individual exclusion keyword, and “no problem” is specified here. In this way, two concept individual keywords and three concept individual exclusion keywords are defined in the concept keyword “not”.

Similarly, as shown in Table 2, as the second keyword, from <keyword k_id = “k-002” type = “concept” name = “to” ”, as the seventh keyword, <keyword
k_id = “k-007” type = “concept” name = “to”> is defined.

Further, as shown in Table 3, <keyword k_id = “k-008” type = “specific” name = “problem”> is defined as the eighth keyword. The keyword type is “specific”, which indicates that this keyword is a specific keyword. The specific keyword “problem” is followed by the definition of one specific individual keyword, and no specific individual excluded keyword is defined. The specific individual keyword is defined as <individual term = "problem"/>. term is the definition of a specific individual keyword, and “problem” is specified here. Furthermore, from the 9th keyword, <keyword k_id = “k-009” type = “concept” name = “to?”>, The 16th keyword, <keyword
k_id = “k-016” type = “specific” name = “film scan”> is defined.

Furthermore, as shown in Table 4, the 17th keyword is <keyword k_id = “k-017” type = “specific” name = “PRT-100”>, and the 23rd keyword is <keyword
k_id = “k-023” type = “specific” name = “SCN-300”> is defined.

このようにして、この例では、キーワード定義部に２５個のキーワードが定義されており、このうち、９個が概念キーワードで、１６個が特定キーワードである。 Furthermore, as shown in Table 5, as the 24th keyword, from <keyword k_id = “k-024” type = “concept” name = “unexpected ~”>, as the 25th keyword, <keyword
k_id = “k-025” type = “specific” name = “abnormal termination”> is defined.

In this way, in this example, 25 keywords are defined in the keyword definition section, of which 9 are conceptual keywords and 16 are specific keywords.

  Next, in the category definition section, as shown in Table 6, <category c_id = “c-001” name = “complaint”> is defined as the first category. c_id is an identifier for uniquely identifying each category, and “c-001” is designated here. The name is a category name, and “complaint” is designated here. This “complaint” category is followed by a definition for referring to the keyword defined in the keyword definition section. At the top, <keyword reference k_id = “k-001” /> is defined. Here, k_id is an identifier for uniquely referring to each defined keyword, and “k-001” is designated. This “k-001” corresponds to the identifier “k-001” of each keyword defined in the keyword definition section. Hereinafter, k_id of the keyword referred to by the category “complaint” is “k-001”, “k-002”, “k-003”, “k-005”, “k-008”, “k-024” , “K-025” is defined. These seven keywords are defined as keywords that mean “complaint” in the meaning recognition rule “rule for user support analysis”. In other words, it is these seven keywords that cause text data to be classified as “complaints”. If any one of these seven keywords is included in certain text data, the text data is classified into the “complaint” category according to the semantic recognition rule.

Similarly, as shown in Table 7, as the second category, <category c_id = “c-002” name = “question”> is the third category, and <category
c_id = “c-003” name = “request”> is the fourth category, <category c_id = “c-004” name = “printer”> is the fifth category, <category
c_id = “c-005” name = “scanner”> is defined as <category c_id = “c-006” name = “new product”> as the sixth category. The definition for referring to the keyword defined in the definition section continues.

  In this way, in this example, six categories are defined in the category definition section, and the keywords referred to in these categories are 7, 2, 2, 8, 6, There are two, 27 in total.

  In the above definition example, there are 25 keywords defined in the keyword definition section, whereas there are a total of 27 keywords referred to in the category of the category definition section. This is because one keyword is referred to from a plurality of categories, for example, the keyword “k-020” is defined in two categories “printer” and “new product”.

  Next, in the application category designation section, the definition for referring to the category defined in the category definition section continues. That is, as shown in Table 8, <reference to category c_id = “c-001” /> is defined at the top. Here, c_id is an identifier for uniquely referring to each defined category, and “c-001” is designated. This “c-001” corresponds to the identifier “c-001” of the individual category defined in the category definition section. In the following, a total of five “c-002”, “c-003”, “c-004”, and “c-006” are defined as c_id of the category referred to by the application category specifying unit. These five categories are defined as categories used in the meaning recognition process in the meaning recognition rule “rule for user support analysis”. On the other hand, categories that are defined in the category definition section but are not defined for reference in the applicable category specification section, that is, “scanners” with c_id “c-005” must not be used for semantic recognition processing. It becomes.

Although not shown in this example, if the system stores keywords and categories in a storage device in a state where they can be referenced from multiple semantic recognition rules, a category once defined can be defined by multiple semantic recognition rules using the applicable category specification function. Since it can be configured freely, it is more convenient for the user.
Next, details of the concept keyword and the specific keyword shown in the meaning recognition rule structure and definition example will be described.

  Regarding keywords, we will explain a mechanism for dealing with various synonyms including notation fluctuations. For example, “sufficient” and “sufficient”, “create” and “create”, “computer” and “computer” are generally synonymous. Such an example can be dealt with by using a universal dictionary. However, for example, a series of keywords “error”, “forced termination”, “freeze”, “abend”, “abort”, “cancel”, and “interrupt” are synonymous, and the keyword name “abnormal termination” is used. Sometimes you want to be able to search. As described above, there is a need to uniquely treat a plurality of keywords as synonyms in a specific field or business. In this example, as in the above example, in order to be able to specify synonyms independently, a mechanism called individual keywords is incorporated into the semantic recognition rule. When a keyword has one or more individual keywords, it represents that these individual keywords are synonymous under the keyword.

  That is, as shown in FIG. 4, one specific keyword 40 has one or more specific individual keywords 40b, and one concept keyword 42 has one or more concept individual keywords 42b. Specifically, three concept individual keywords are designated for the fourth keyword “Please” shown in Table 2. Similarly, eight specific individual keywords are designated for the 25th keyword “abnormal end” shown in Table 5. Here, as a synonym of the keyword “please please”, “please please”, “can you give me” and “can you give me” are defined. Similarly, "abnormal termination", "error", "forced termination", "freeze", "abend", "abort", "cancel" and "suspend" are defined as synonyms for the keyword "abnormal termination". .

  When matching with text data using the keyword as described above, it matches the defined keyword but has a different meaning, so there may be a need to exclude it from the extraction target. For example, in the meaning recognition rule for user support analysis, the first keyword “not” shown in Table 1 includes three individual excluded keywords “no problem”, “not bad”, and “no mistake”. It is specified. The negative expression “not” is valid as a keyword meaning “complaint” in many cases, but there are exceptions. That is, the expressions “no problem”, “not bad”, and “no doubt” are inappropriate for meaning “complaint”.

  In this example, in order to cope with such a situation, as described above, the mechanism of individual exclusion keywords is incorporated in the semantic recognition rule. When a character string included in text data matches an individual keyword of a certain keyword, but also matches an individual excluded keyword of the same keyword, the character string is excluded from extraction targets.

  Hereinafter, a method for realizing individual keywords will be described. The concept individual keyword is defined for abstract matching with a character string included in text data. For example, in the case of the first keyword “not” in Table 1, since the abstractly expressed portion “to” is included, it is not possible to simply compare only character strings.

日本語における連続した複数の形態素の並びがあったとして、この複数の形態素の１つ以上が抽象形態素の場合、この連続した形態素の並びを概念個別キーワードとする。一方、１個以上の連続する形態素が全て具象形態素の場合、この連続する形態素の並びを特定個別キーワードとする。また、概念個別除外キーワードと特定個別除外キーワードについても同様である。 Based on this, in this example, the text data is analyzed into Japanese morphemes and divided into three attributes, “basic name”, “part of speech”, and “utilization”, and the concept individual keyword and the specific individual keyword shown below. Is realized. For example, when the morpheme of the expression “unreadable” is analyzed, as shown in Table 9, for the “read” part, the basic name is “read”, the part of speech is “verb”, and the utilization is “unformed” . Further, as shown in Table 10, for the part of “None”, the basic name is “None”, the part of speech is “auxiliary verb”, and the utilization is “basic form”. A morpheme whose contents of these morpheme attributes are all specified is called a concrete morpheme. A morpheme in which part or all of the content of the morpheme attribute is not specified is called an abstract morpheme.

If there is a sequence of a plurality of continuous morphemes in Japanese and one or more of the plurality of morphemes is an abstract morpheme, this sequence of morphemes is used as a concept individual keyword. On the other hand, when one or more continuous morphemes are all concrete morphemes, this sequence of continuous morphemes is set as a specific individual keyword. The same applies to the concept individual excluded keyword and the specific individual excluded keyword.

また、表１１の形態素属性を全て特定されないようにすれば、表１３に示すようになる。表１３には、全ての形態素が該当する。例えば、表１３と表１２の形態素の連続による「〜ない」という概念個別キーワードを設定すれば、前記動詞の「読めない」以外にも、格助詞の「応答がない」、形容詞の「正しくない」、名詞の「問題なし」なども当該概念個別キーワードに該当する。

このようにして、日本語形態素属性を使って概念個別キーワードを実現することができるが、形態素属性の内容の指定方法については、特別の方法で実現する必要はないため、ここではその説明を省略する。 For the individual keyword “not”, for example, when there is an expression “unreadable”, two morphemes “read” and “not” are arranged, as shown in Table 9 and Table 10. Tables 11 and 12 show some abstractions of the attributes of these morphemes. The attribute corresponding to the basic name in Table 11 and the attribute corresponding to utilization in Table 12 are “-”, indicating that they are not specified and are abstracted. Table 11 is an abstraction of the attribute of the basic name in the morpheme, which is the verbal verb “Read” in Table 9, and the verbal form is the same regardless of the content of the basic name. “ Does not move ”, “ does not move ”, and so on. Therefore, if the morphemes of Table 11 and Table 12 are continued and the morpheme whose base name is abstracted is expressed by the special character “to”, the concept individual keyword “to not” can be realized.

Further, if all the morpheme attributes in Table 11 are not specified, the result is as shown in Table 13. Table 13 corresponds to all morphemes. For example, by setting the concept individual keyword "no ~" by the morpheme of a series of Table 13 and Table 12, in addition to the "unreadable" of the verb, "there is no response" of the case particle, "not correct adjectives ", No problem ", etc., also fall under the concept individual keyword.

In this way, it is possible to realize concept individual keywords using Japanese morpheme attributes, but the method for specifying the contents of morpheme attributes does not need to be implemented in a special way, so the explanation is omitted here. To do.

  Next, how the user actually defines the semantic recognition rule using the semantic recognition rule definition tool 20 shown in FIG. 2 will be described with reference to FIGS. When the user requests the text data analysis system using the operation buttons to use the meaning recognition rule definition tool 20, the system calls the rule name designation function 20d and displays the meaning recognition rule name designation dialog on the display device 14. FIG. 5 is a display example thereof. Here, the user designates the meaning recognition rule name “rule for user support analysis” using the input device 16. When the “Next” button is selected, the specified rule name is stored in the storage device 12 and the process proceeds to the next process. At this point, the rule name 32 in the meaning recognition rule 30 shown in FIG. 4 is completed.

  Next, the text data analysis system calls the keyword definition function 20a and displays a keyword definition dialog on the display device 14. FIG. 6 shows an example of the display. In the keyword definition function 20a, when the user designates a specific keyword or conceptual keyword using the input device 16 and selects an “add” button, each keyword is defined in the semantic recognition rule. Here, the user selects “specific” (●) and designates the keyword “printer”. In order to specify a synonym of a keyword, a concept individual excluded keyword, or a specific individual excluded keyword, the “details” button is selected. After the selection, a keyword detail designation dialog is displayed on the display device 14 so that it can be designated. However, since it is not a special method, its description is omitted here. When the “next” button is selected, the keyword specified by the keyword definition function 20a is stored in the storage device 12, and the process proceeds to the next process. At this point, the keyword definition unit 34 in the meaning recognition rule 30 shown in FIG. 4 is completed.

  When the keyword definition is completed, the text data analysis system using the operation buttons calls the category definition function 20b and displays the category definition dialog on the display device 14. FIG. 7 is a display example thereof. Here, the user designates the category of “request” from the input device 16. In addition, a list of keywords defined by the keyword definition function 20a so far is displayed with the heading “designate keywords to be associated”. By selecting one or more arbitrary keywords from the list, the keywords are associated with the category of the designated name. In this example, the keywords “to please” and “want to want” are selected to be associated with the category named “request”. When the “add” button is selected, the category of the designated name and the keyword associated with the category (“reference to the defined keyword” shown in the structure of the semantic recognition rule in FIG. 4) are added to the semantic recognition rule. Defined. When the “Next” button is selected, the definition information is saved in the storage device 12 and the process proceeds to the next process. At this point, the category definition unit 36 in the semantic recognition rule 30 shown in FIG. 4 is completed.

  When the category definition is completed, the meaning recognition rule is completed in the storage device 12. Note that the application category designation function simply selects a reference to a predefined category, and the description thereof is omitted here. The text data analysis system using operation buttons displays a file save dialog on the display device 14 in order to save the completed meaning recognition rule in the meaning recognition rule file. FIG. 8 shows an example of the display. Here, the user designates a file “file: / rules / user_support_analysis”. When the “Done” button is selected, the text data analysis system using the operation button stores the contents defined by each function of the semantic recognition rule in the designated semantic recognition rule file, and ends the semantic recognition rule definition process. .

  The following describes how the user actually uses the semantic recognition tool. When the user requests the text data analysis system using the operation buttons to use the meaning recognition tool 22 shown in FIG. 2, the system calls the analysis target text data specifying function 22a and the meaning recognition rule specifying function 22b, and the meaning recognition tool. A dialog is displayed on the display device 14. FIG. 9 is a display example.

  Here, the user designates “file: / text / user_support /” as the location where the text data to be analyzed shown in FIG. 3 is stored. Further, the meaning recognition rule file shown in FIG. 8 is designated as “file: / rules / user_support_analysis”. Thus, the specification for performing the semantic recognition process on the text data stored in file: / text / user_support / according to the semantic recognition rules stored in file: / rules / user_support_analysis is completed.

  When the “semantic recognition start” button shown in FIG. 9 is selected, the text data analysis system using the operation buttons calls the buttoned source data creation function 22c to start the semantic recognition process. Based on the result, buttoned source data 50 shown in FIG. 10 is created and stored in the storage device 12. In the storage device 12, it is stored in an internal file called a buttoned source data file. When the saving is completed, the meaning recognition process is terminated.

  As shown in FIG. 10, the buttoned source data 50 includes a meaning recognition rule file name 52, an analysis target text data storage location 54, and a meaning recognition result table 56. The meaning recognition result table 56 is constituted by a set of result records 58 having at least three fields of a category 58a, a keyword 58b, and a text data file name 58c. The category and the keyword are represented by c_id and k_id, which are identifiers. The text data file name 58c is the file name shown in the legend: text unit of FIG. “1” in FIG. 10 is one, and “**” is zero or more.

In the following, an example of creating the button source data 50 by the designation in FIG. 9 in the semantic recognition tool will be described.
Table 14 is an example in which the meaning of all text data from “text A” to “text M” in the text data to be analyzed shown in FIG. 3 is recognized, and buttoned source data is created based on the result. .

  Here, an example in which the content of text data is analyzed using the definition examples of the semantic recognition rules shown in Tables 1 to 8 by paying attention to “Text A”. The content of the text data of “Text A” is “I cannot print after updating the printer driver of Print King 100. Please tell me how to deal with it”. When keywords defined in the definition example of the semantic recognition rule are extracted from the list, “~ No”, “~ Take”, “~ Please”, “Printer”, “Print” and “PRT-100” It will be six. Here, among the extracted keywords, for “print” and “PRT-100”, the content of “text A” includes the character string “printing 100”. In the definition example, “printing”, which is an individual exclusion keyword for the keyword “printing”, is defined, so that “printing” is defined as an individual keyword for the keyword, but the character string is “printing”. It will not be extracted. In addition, in the definition example of the meaning recognition rule in Table 4, the character string is extracted as “PRT-100” by defining “Print King 100” which is an individual keyword of the keyword “PRT-100”. It will be.

  First, for “˜not”, k_id = “k-001” type = “concept” is defined in the keyword definition part of the above definition example, and the reference to the keyword in the category definition part is k_id = “k-001”. The category “” is defined as c_id = “c-001” category name = “complaint”. From this, it is understood that “text A” including “not” belongs to the category “complaint”, the identifier is “c-001”, and the identifier of the related keyword is “k-001”. When the result record 58 of the meaning recognition result table 56 of the button source data 50 shown in FIG. 10 is created based on this result, as shown in <Result record> at the top of <Result table> in Table 14, category = “C-001”, keyword = “k-001”, doc = “text A”. Here, doc refers to the legend of FIG. 3: file name in text units.

  Next, for “to end”, k_id = “k-003” type = “concept” is defined in the keyword definition part, and the reference to the keyword is k_id = “k-003” in the category definition part. A certain category is defined as c_id = “c-001” category name = “complaint”. From this, “Text A” including “to end” belongs to the category of “complaint” like “to not”, the identifier is “c-001”, and the identifier of the related keyword is “k-”. 003 ”. When the result record 58 of the semantic recognition result table 56 of the button source data 50 shown in FIG. 10 is created based on this result, as shown in the sixth <result record> from the top of the <result table> in Table 14. , Category = “c-001”, keyword = “k-003”, doc = “text A”.

In the following, button source data is created in the same way for the keywords after “to please”. In this way, the meaning of the text data is recognized not only for “text A” but also for text units after “text B” by the same method.
In this way, the target text data and semantic recognition rules specified in FIG. 9 are used to recognize the meaning of the contents of the text data in all text units included in the target text data, and the buttoned source data shown in FIG. According to the structure of 50, it is output to an internal file called a buttoned source data file and stored in the storage device 12.

  With buttoned source data, you can create an operating environment with semantic buttons. When the user requests the text data analysis system using operation buttons to use the semantic button automatic generation tool 24 shown in FIG. 2, the system calls the semantic button automatic generation function and reads the button source data from the storage device 12. , Meaning buttons are automatically generated and displayed on the display device 14.

  That is, the button class generation function 24a of the semantic button automatic generation tool 24 extracts a category from the button source data and generates a button class based on the extracted category. A button class is one of the elements constituting a semantic button and is generated for each category. Individual buttons belonging to the same category are collected and displayed in one button class. An individual button is one of the elements constituting a semantic button and is generated for each keyword. Collect and display for each button class corresponding to the category. When an individual button is selected, attention is paid to other keywords included in the text data including the keyword corresponding to the individual button, and the individual buttons of the button class corresponding to the category to which the keyword belongs are limited to those corresponding to the keyword. To redisplay.

  When buttoned source data shown in Table 14 above is used, categories with identifiers "c-001", "c-002", "c-003", and "c-004" correspond to button classes Then, according to the category definition in the definition example of the meaning recognition rule, a button class is generated by associating each of the category names “complaint”, “question”, “request”, and “printer” with the button class name. Regardless of the contents of the text data, the doc (file name in text unit) of the result table (record) of the buttoned source data is always the target of the reference button class. There are two types of button classes for semantic buttons: analysis and reference. The former is used to analyze text data by selecting individual buttons corresponding to keywords having important meanings. The latter is used to select the individual button corresponding to the unique data value and retrieve relevant data.

  Depending on the usage, if there are fields for reference other than the text data to be analyzed, such as “author” and “creation date / time” in the result table (record) of the buttoned source data, they are also treated as categories, and the button class for reference can do.

  In addition, in the definition example of the semantic recognition rule applied to the buttoned source data, a category with the name “new product” (c_id = “c-006”) is defined and specified as an applied category. As a result of recognition, there was no text data corresponding to this category. In this case, since the result record having the category “c-006” is not output in the button source data result table, the button class corresponding to this category is naturally not generated.

  The individual button generation function 24b of the semantic button automatic generation tool 24 extracts keywords from the button source data and generates individual buttons based on the keywords. The individual buttons depend on the button class, but the method of generating the individual buttons differs depending on whether the button belongs to the analysis button class or the reference button class.

A method for generating individual buttons belonging to the button class for analysis will be described. Focusing on one category existing in the result table shown in Table 14, only the result records that match the identifier of the category are narrowed down. A unique keyword identifier is extracted from the narrowed result record, and an individual button belonging to the button class corresponding to the category is generated. Focusing on the category identifier “c-001” (category name “complaint”) in the buttoned source data shown in Table 14 above, it is possible to narrow down to 11 result records shown in Table 15.

  The identifiers for the unique keywords present in these result records are “k-001”, “k-002”, “k-003”, “k-005”, “k-025”, respectively ”,“ ˜zu ”,“ to end ”,“ ˜funny ”, and“ abnormal termination ”. Based on these five keywords, an individual button belonging to the button class “complaint” (for analysis) corresponding to the category identifier “c-001” is generated.

  The method for generating the reference button class and the individual buttons is shown below. This method is based on the invention described in Patent Document 1. By combining the operation button according to the present invention and the operation button generation method according to the invention described in Patent Document 1, a more useful operation button can be realized.

  In order to generate a button class based on the reference field existing in the result table shown in Table 14, the field values of all result records are extracted. Based on the unique value, an individual button belonging to the reference button class corresponding to the field is generated. In the example of creating the button source data described above, paying attention to the doc (file name in text unit) of the reference field, “Text A”, “Text B”, “Text C”, “Text D”, “Text E”, “Text F”, “Text G”, “Text H”, “Text I”, “Text J”, “Text K”, “Text L” and “Text M” is extracted, and an individual button of a reference button class (named “target”) is generated based on these.

  The category field in the semantic recognition result table is used for generating the analysis button class as described above, but it can also be used to generate the reference button class by treating it as a reference field. . In this example, focus on the category field as a reference field, and from all the result records, the unique value of the field is “c-001”, “c-002”, “c-003”, “c- 004 "is extracted, the corresponding category names are extracted from the category definition section in the meaning recognition rule definition example, and based on these, individual buttons of the reference button class (named" main classification ") are generated. . Since the unique value of the field corresponds to one or more meaning recognition result records, the individual button name belonging to the button class is determined by adding “group” after the category name. As described above, the individual buttons belonging to the button class are “complaint group”, “question group”, “request group”, and “printer group”.

  When the semantic button is generated from the buttoned source data by the semantic button automatic generation tool, the text data analysis system by the operation button displays the analysis tool dialog by the semantic button on the display device 14. FIG. 11 is a display example.

  As shown in FIG. 2, the semantic button analysis tool 26 includes a semantic button parallel relation display function 26a and a text data search and content display function 26b. The dialog shown in FIG. 11 is displayed by the meaning button parallel relation display function 26a. “Main classification”, “complaint”, “question”, “request”, “printer”, and “target” in FIG. 11 are button classes displayed in parallel. Among these, the button class from “complaint” to “printer” is an analysis button class corresponding to the category of the result record, “main classification” is a reference button class corresponding to the category, and “target” "Is a button class for reference corresponding to doc (file name in text unit). The “complaint group”, “question group”, “request group”, and “printer group” displayed under the button class “main classification” are individual buttons belonging to the “main classification”. Similarly, individual buttons displayed under each button class from “complaint” to “target” are individual buttons belonging to each button class. Individual buttons belonging to each button class are associated with individual buttons belonging to other button classes through analysis target text data, and are selected to narrow down the text data.

  When the individual buttons of the button class “target” are narrowed down to one by the selection of the individual buttons, that is, when the file name of the text unit is narrowed down to one by the selection of the individual buttons, the text data search and content display function 26b. Thus, the file is read out and the contents thereof are displayed in the “text content being selected” positioned below the dialog shown in FIG. FIG. 11 shows a state in which no individual button is selected, and since the “target” individual button is not narrowed down to one, the text content is not displayed.

  When the individual button “text A” belonging to the button class “target” is selected in the state shown in FIG. 11, the display of the entire semantic button is updated as shown in FIG. “Text A” is shown in a bold frame). Specifically, only the individual buttons related to “text A”, that is, the individual buttons of the button class corresponding to the keyword included in the text are redisplayed.

  As shown in FIG. 12, “text A” is classified into three categories, “complaint”, “request”, and “printer”. The keywords “to not” and “to end” belonging to the category “complaint” are extracted, and individual buttons corresponding to these are displayed. Similarly, a keyword “to please” belonging to “request” of the category is extracted, and an individual button corresponding to this is displayed. Further, keywords “printer”, “print”, and “PRT-100” belonging to the category “printer” are extracted, and individual buttons corresponding to these are displayed. Since the individual buttons “question group” from “main classification” are hidden, and at the same time, all the individual buttons belonging to the button class “question” are hidden, the “text A” includes “ It can be seen that the keyword belonging to “question” is not included.

  Since the text unit is narrowed down to one file name “text A”, the contents of “text A” are displayed by the text data search and content display function 26b. In the display contents of FIG. 12, the underlined bold part is a character string corresponding to the extracted keyword.

  In the state shown in FIG. 11, when the individual button “not” of the button class “complaint” is selected, the display of the whole semantic button is updated as shown in FIG. 13 by the semantic button parallel relation display function 26a. Specifically, only individual buttons related to “not”, that is, only individual buttons of button classes corresponding to other keywords included in “text A” and “text J” including the keyword “not”. It will be displayed again.

  In FIG. 13, it is understood that “text” and “text J” are included in the “object” referring to the text data that includes the keyword “not” belonging to the category “complaint”. In addition, these texts include the keywords “to please” belonging to the category “request” and the keywords “printer”, “print” and “PRT-100” belonging to the category “printer”. It can be seen that the keywords belonging to the category “Question” are not included.

  When the individual button “abnormal end” of the button class “complaint” is selected in the state shown in FIG. 13, the display of the whole semantic button is updated as shown in FIG. 14 by the semantic button parallel relation display function 26a. Specifically, button classes corresponding to other keywords included in the “text A”, “text J”, “text L”, and “text M” including the keywords “to not” or “abnormal end”. Only individual buttons are redisplayed.

  In FIG. 14, the keywords “to not” or “abnormal termination” belonging to the category “complaint” include “text A”, “text J”, “text L” and “object” referring to text data. It turns out that it is "text M". In addition, these texts include the keywords “~ Please” and “~ I want” belonging to the category “Request”, and “Printer”, “Print” and “PRT-100” belonging to the category “Printer”. It is understood that the keyword belonging to the “question” of the category is not included.

  Furthermore, when the individual button “question group” of the button class “main classification” is selected in the state shown in FIG. 11, the display of the whole semantic button is updated as shown in FIG. 15 by the semantic button parallel relation display function 26a. Specifically, all of the keywords belonging to the category “question”, that is, “text B”, “text C”, “text G”, and “text H” including the keywords “˜ka” or “˜?” Only the individual buttons of the button class corresponding to the other keywords included are redisplayed.

  In FIG. 15, “question group” corresponds to “a question group” (that is, includes any one of the keywords “˜ka” and “˜?”). C ”,“ Text G ”, and“ Text H ”. These texts include at least one of the keywords“ printer ”,“ ink ”, and“ PRT-200 ”that belong to the category“ printer ”. It can be seen that is included. It can also be seen that the keywords belonging to the categories “complaint” and “request” are not included.

  Further, when the individual button “ink” of the button class “printer” is selected in the state shown in FIG. 15, the display of the whole semantic button is updated as shown in FIG. 16 by the semantic button parallel relation display function 26a. Specifically, only the individual buttons “text C” and “text H” corresponding to the “question group” and including the keyword “ink” are displayed again.

In FIG. 16, the keyword “ink” that corresponds to the “question group” and belongs to the category “printer” includes “text C” and “text H” in “object” that refers to text data. Yes, it can be seen that these texts do not include keywords belonging to the categories “complaint” and “request”.
In the example of the analysis tool using the semantic buttons described above, when multiple individual buttons are selected between different button classes, the narrowing-down conditions by selection are combined by logical product (AND), and individual buttons within the same button class. When a plurality of are selected, the narrowing-down conditions by selection are combined by a logical sum (OR).

  In addition to the above, the logical operation of the filtering condition when selecting multiple buttons is logical OR (OR) between different button classes, logical product (AND) in the same button class, both logical product (AND), both A combination of logical OR (OR) is also conceivable, but of course, all of these can be realized.

  As described above, according to this example, an operation button (meaning button) is automatically generated based on various meanings of text data, and this button is selected even by an end user who has little computer experience and does not have expertise. Just search a large amount of text data from various viewpoints corresponding to an arbitrary category and keyword while dynamically narrowing down the hierarchy, and redisplay multiple button classes and individual buttons in parallel. You can easily search while looking at the correlation with other categories of keywords.

It is a figure which shows the outline | summary of the computer system for implementing this invention. It is a figure which shows the software structure of the text data analysis system by an operation button. It is a figure which shows the example of the text data used as an analysis object. It is a figure which shows the structure of a meaning recognition rule. It is a figure which shows the example of a meaning recognition rule name designation | designated dialog. It is a figure which shows the example of a keyword definition dialog. It is a figure which shows the example of a category definition dialog. It is a figure which shows the example of a file save dialog. It is a figure which shows the example of a meaning recognition dialog. It is a figure which shows the structure of buttoned source data. It is a figure which shows the analysis tool dialog by a meaning button. It is a figure which shows the analysis tool dialog by a semantic button when the individual button "text A" of button class "target" is selected and redisplayed in the state shown in FIG. It is a figure which shows the analysis tool dialog by a semantic button when the individual button "not" of button class "complaint" is selected and redisplayed in the state shown in FIG. It is a figure which shows the analysis tool dialog by a semantic button when the individual button "abnormal end" of button class "complaint" is selected and redisplayed in the state shown in FIG. It is a figure which shows the analysis tool dialog by a semantic button when the individual button "question group" of button class "main classification" is selected and redisplayed in the state shown in FIG. FIG. 16 is a diagram showing an analysis tool dialog with semantic buttons when the individual button “ink” of the button class “printer” is selected and displayed again in the state shown in FIG. 15.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Central processing unit 12 Storage device 14 Display device 16 Input device 20 Meaning recognition rule definition tool 20a Keyword definition function 20b Category definition function 20c Applicable category specification function 20d Rule name specification function 22 Meaning recognition tool 22a Analysis target text data specification function 22b Meaning Recognition rule specification function 22c Buttoned source data creation function 24 Meaning button automatic generation tool 24a Button class generation function 24b Individual button generation function 26 Meaning button analysis tool 26a Meaning button parallel relation display function 26b Text data search and content display function 30 Meaning Recognition rule 32 Rule name 34 Keyword definition unit 36 Category definition unit 38 Application category specification unit 40 Specific keyword 40a Specific keyword name 40b Specific individual keyword 40c Specific individual exclusion Keyword 42 Concept keyword 42a Concept keyword name 42b Concept individual keyword 42c Concept individual exclusion keyword 44 Category 44a Category name 44b Reference to a defined keyword 46 Reference to a defined category 50 Buttoned source data 52 Semantic recognition rule file name 54 Analysis target Text data storage location 56 Semantic recognition result table 58 Result record 58a Category 58b Keyword 58c Text data file name

Claims (5)

Computer processing of text data generated by processing of a computer programmed with a computer processing operation button for retrieving arbitrary text data from a plurality of text data stored in a storage device with a file name including a character string Operation button generation method,
Defining a category and a keyword that is an expression element for matching with a character string in the text data,
Search for text data in which the keyword is included in a character string in the text data,
The category, the keywords, and the text data or et al., Which it has been retrieved that contain the keyword, category field, have a three field of keyword field and text data file name field, in each field, the Generate buttonized source data with a semantic recognition result table that is a set of result records that correspond one-to-one with the category, the keyword, and the file name of the text data searched to contain the keyword. And
Analysis button CLASS in correspondence with the field value of the category field of the button of the source data, individual buttons respectively generating belonging to the button class for the analysis in correspondence to the field values of the keywords field,
The browse button CLASS from the text data file name field of the button of the source data, the text data file name field of the field value is associated with an individual button belonging to the button class for the reference product, respectively,
An operation button generation method for computer processing of text data, wherein the button class and the individual buttons are displayed on a display device. The analysis button class and the reference button class are displayed in parallel on the display device together with the individual buttons, and when any individual button belonging to one analysis button class is selected, the selected individual button corresponds to the selected individual button The result record including the keyword corresponding to the field value of the keyword field and at least one of the other keywords included in the text data including the keyword is extracted from the semantic recognition result table, and the extracted result record 2. The operation button generating method for computer processing of text data according to claim 1, wherein the individual buttons belonging to the analysis button class and the reference button class are generated and redisplayed. Meaning the result record including all the keywords respectively corresponding to the field values of the keyword field corresponding to the selected individual button by selecting any individual button of any button class for analysis in any order 3. The method according to claim 1 , further comprising: extracting from the recognition result table, and generating and redisplaying the individual buttons belonging to the analysis button class and the reference button class based on the extracted result record. An operation button generation method for computer processing of the described text data. Define the keyword,
By defining separately a reference to the predefined keywords for each of the categories, to define the category while associating the categories and keywords each other,
Extracting a keyword that matches the keyword from the text data and associating it with a category in which a reference to the keyword is defined;
4. The operation button generation method for computer processing of text data according to claim 1, wherein the meaning recognition result table is generated based on the associated result. The keywords are categorized and defined as a specific keyword composed of a specific character string and a concept keyword composed of a character string including an abstracted portion. In the specific keyword, the specific character string is defined as the concept. The keyword is used to match a character string excluding the abstracted portion from a character string including the abstracted portion with a character string in the text data. 5. An operation button generation method for computer processing of text data according to any one of 4 above.

Images