JP2012234282A - Request document analysis system, method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve ambiguity of a request related document for use in an upstream process of an information system configuration.SOLUTION: An order of priority of ambiguity is set for ambiguity points included in the request related document for use in the upstream process of the information system configuration, based on an evaluation criterion unique for the request related document. The ambiguity points are narrowed to those that are highly likely to cause erroneous modification and presented so as to improve the ambiguity of the request related document for use in the upstream process of the information system configuration.

Description

  The present invention relates to a requirement document analysis system, method and program, and in particular, a request related document such as a request for proposal issued by a customer or a proposal or specification provided by a developer in an upstream process of information system construction. The present invention relates to a requirement document analysis system, method and program.

  An example of prior art relating to a requested document analysis system, method, and program is described in Patent Document 1 as a “translation device”. The translation apparatus disclosed in Patent Document 1 includes an input unit, a translation unit, a storage unit, and an output unit. The translation apparatus having such a configuration operates as follows.

  That is, the translation unit detects the break position of the original text input from the input unit as the interruption position. Further, the translation unit searches for a candidate for analysis by parsing a word string ahead of the interruption position in the original text based on an analysis rule registered in the storage unit. And a translation part adds the necessity level of the inquiry based on the inquiry sentence with respect to an operator, and the applied analysis rule about the place with a some analysis candidate. Further, the display unit displays an inquiry sentence for an analysis candidate having a necessity level higher than the inquiry frequency level set. The translation device discards the analysis candidate when the response to the displayed query input from the input unit is the discard of the analysis candidate, and when there is ambiguity in the dependency of the sentence, The ambiguity is removed by dialogue with the operator, and the ambiguous analysis results are uniquely narrowed down.

  Further, another example of the related art relating to the requested document analysis system, method and program is described in Patent Document 2 as “Software Specification Description Method”. In the software specification description method disclosed in Patent Document 2, the basic vocabulary and detailed vocabulary expressing the components of the model that abstracts the target system are included in a pseudo-code sentence that describes the overall processing flow. The specification is described using the basic vocabulary and the detailed vocabulary associated with each other by using an assignment table that describes the state of assignment of specific objects to the detailed vocabulary. Also, in this software specification description method, a basic process description sentence is described using a basic vocabulary that expresses the components of the model that abstracts the target system, and this basic process description sentence is used in the pseudo code sentence. The specification is described. In addition, this software specification description method creates a verb vocabulary relation table that defines the relationship between the processing vocabulary meaning of the verb vocabulary and case particles and fills in the corresponding basic vocabulary, and creates a machine for pseudo-code sentences written by humans. Is performing a dynamic check. With such a configuration, the ambiguous syntax and the possibility of erroneous specification description are presented to the user to promote correction.

Japanese Patent Application Laid-Open No. 08-297474 JP 2000-132382 A

  The first problem of the prior art is that if the vague point presentation method according to the prior art is applied to improve the ambiguity of requirements-related documents used in the upstream process of information system construction, the burden on the worker will increase. That is. The reason is that the requirement-related document used in the upstream process of information system construction is premised on being read carefully by humans, and humans can judge points where the degree of ambiguity is relatively weak. For this reason, pointing out ambiguous points indiscriminately and requesting corrections only adds to the cumbersome work, and it is necessary to present only those points that have a high risk of misinterpretation due to assumptions, etc. It is to become.

  Furthermore, the second problem of the prior art is that the vague point presentation method according to the prior art is not suitable for improving the ambiguity of requirement-related documents used in the upstream process of information system construction. The reason for this is that requirements-related documents used in the upstream process of information system construction are supposed to have the meaning and usage of unique words in requirements-related documents. This is because there is no narrowing down and prioritization based on evaluation criteria specific to requirement-related documents is not performed.

  An object of the present invention is to provide a requirement document analysis system, method, and program for improving the ambiguity of requirement-related documents used in the upstream process of information system construction.

  A requested document analysis system according to the present invention is a system for analyzing a requested document, which applies morphological analysis and syntax analysis to each sentence of a target document and extracts a word set based on a predetermined extraction rule A set extraction unit; for each word set, based on a predetermined basic axis setting rule, extract a basic word from each word included in the word set, and classify the word set into a basic word and peripheral words that are other words A basic word extraction unit; accumulates semantic information of words, synonyms, synonyms, etc., and retrieves and responds to inquiries about specific words by retrieving semantic information related to the meaning and usage of words A word database; a group of word sets having the same basic word from all word sets, and the above word similarity data for each peripheral word of the word set group A word similarity calculation unit that calculates the similarity between words based on the semantic information that has been queried and sent back; and between adjacent words for each word set group having the same basic word in the document A word classification unit that classifies peripheral words based on similarity and classifies the peripheral words; from the classification results performed by the word classification unit, an extraction criterion for each word set group having the same basic word in the document A non-mainstream cluster is extracted as a non-mainstream cluster as a word usage, and an ambiguous point extraction unit that extracts a combination of a peripheral word and a base word belonging to the non-mainstream cluster as an ambiguous point that is assumed to be a potential mistake An ambiguous point output unit that outputs an ambiguous point extracted for each word set group having the same basic word in the document by the ambiguous point extracting unit; That.

  The effect of the present invention is to easily grasp and accurately understand ambiguous points that are likely to be misinterpreted in a requirement document that is assumed to have a meaning and usage of unique words in the requirement-related document. It is possible to reduce the rework in subsequent processes. The reason for this is to prioritize and present the ambiguity points based on the evaluation criteria specific to the request-related document.

It is a block diagram which shows the structure of the request document analysis system which concerns on the 1st Embodiment of this invention. It is a sequence diagram which shows operation | movement of the request document analysis system shown in FIG. It is a block diagram which shows the structure of the request document analysis system which concerns on the 2nd Embodiment of this invention. It is a sequence diagram which shows operation | movement of the request document analysis system shown in FIG. It is a block diagram which shows the structure of the request document analysis system which concerns on 1st Example of this invention. It is an example of a part of result of calculating similarity Fa between predicates. It is an example of the dendrogram Ha created based on the similarity Fa between predicates. This is an example of extracting a predicate Vax belonging to a non-mainstream cluster from the dendrogram Ha. It is a block diagram which shows the structure of the request document analysis system which concerns on 2nd Example of this invention. This is an example in which the general-purpose degree J of words and the appropriate distance between clusters are mapped.

[Embodiment 1]
First, a first embodiment of the present invention will be described in detail with reference to the drawings.

  FIG. 1 is a block diagram showing the configuration of a requested document analysis system 100 according to the first embodiment of the present invention.

  Referring to FIG. 1, a requested document analysis system 100 according to the first embodiment of the present invention is basically a system comprising an electronic device or a server and an electronic device, and an information communication network such as the Internet for connecting them together. Among them, at least a word set extraction unit 10, a base word extraction unit 20, a word similarity calculation unit 30, a word similarity database 110, a word classification unit 40, an ambiguous point extraction unit 50, and an ambiguous point output Part 60.

  The illustrated requirement document analysis system 100 is a requirement document analysis system for requirement-related documents such as a proposal request document issued from a customer or a proposal document or specification provided from a developer in an upstream process of information system construction.

  When the required document analysis system is configured by an electronic device, the required document analysis system 100 can be realized by a computer that operates under program control. Although not shown, this type of computer, as is well known, includes an input device for inputting data, a data processing device, an output device for outputting processing results in the data processing device, and an auxiliary memory serving as various databases. Device. Then, the data processing device stores data in a read-only memory (ROM) that stores a program, a random access memory (RAM) that is used as a work area that temporarily stores data, and a program stored in the ROM. It consists of a central processing unit (CPU) that processes stored data.

  In this case, the data processing device functions as the word set extraction unit 10, the base word extraction unit 20, the word similarity calculation unit 30, the word classification unit 40, and the ambiguous point extraction unit 50, and the auxiliary storage device functions as the word similarity database 110. The output device functions as the ambiguous point output unit 60.

  Next, the operation of each component constituting the request document analysis system 100 will be described.

  The word set extraction unit 10 receives an input of a target document or document group. The word set extraction unit 10 applies morphological analysis and syntax analysis to all sentences constituting a document or a document group, and extracts a word set based on a predetermined extraction rule. Here, any method may be used as the “extraction rule” as long as it is a rule reflecting usage characteristics between words constituting a sentence. For example, if attention is given to co-occurrence as a feature, the extraction rule corresponds to a rule such as extracting a combination of words that co-occur within an arbitrary range of a document such as a sentence as a word set. Further, if attention is paid to the dependency relationship between words, the extraction rule corresponds to a rule of extracting a combination of words having a dependency relationship as a word set.

  The base word extraction unit 20 extracts a base word from each word included in the word set based on a predetermined base setting rule for each word set. Then, the basic word extraction unit 20 classifies the words in the word set into basic words and peripheral words that are other words. Here, as the “base axis setting rule”, any method may be used as long as it is a rule for extracting a central word of a sentence. For example, the basic setting rule is a rule in which a word that is relatively frequently used in a general Web or a document among the parts of speech that make sense by itself, such as nouns, adjectives, and verbs, is used as a basic word. Applies. Or, in the case of a word set extracted based on the dependency relationship, the basic setting rule is a rule in which the related word or subject is the basic word, and the related word or predicate, the particle connecting the subject and the predicate is a peripheral word, etc. Applies.

  The word similarity database 110 collects and accumulates semantic classification of words and semantic information such as synonyms and synonyms, and searches for semantic information related to the meaning and usage of a word in response to an inquiry about a specific word. The database that responds. The word similarity database 110 categorizes words by upper / lower relations, partial / whole relations, synonym relations, synonym relations, etc., and organizes each word in a structured thesaurus or general Web or document text. A co-occurrence word count table that counts the types and the number of co-occurrence words that co-occur with.

  The word similarity calculation unit 30 collects word sets having the same basic word from all word sets included in the document, and stores semantic information in the word similarity database 110 for each peripheral word in the word set group. And the similarity between neighboring words is calculated based on the returned semantic information. When the word similarity database 110 is a thesaurus, the method of calculating the similarity between neighboring words may be a method using the depth of the hierarchy in which words are a common concept as an index of dissimilarity. Alternatively, as a method of calculating similarity between neighboring words, the type and number of co-occurrence words that co-occur with each neighboring word in a sentence in a general Web or document are extracted from the co-occurrence word summary table, and the neighboring words A method may be used in which the overlapping rate of co-occurrence words is used as an index of similarity.

  The word classification unit 40 classifies the peripheral words by clustering the peripheral words based on the similarity between the peripheral words for each word set group having the same basic word in the document. The clustering method may be a general method. For example, hierarchical clustering using a dendrogram or the like may be applied as a clustering method. Alternatively, as a clustering method, non-hierarchical methods such as a k-means method and a Fussy-c-means method are used for position information of virtual neighboring words derived by using similarity between neighboring words and a monotonically decreasing index as a distance. Clustering may be applied. Alternatively, when the word similarity database 110 is a thesaurus, any one of the major classification, middle classification, and minor classification to which peripheral words in the thesaurus belong may be used as the classification.

  The ambiguous point extraction unit 50 determines, as a non-mainstream cluster, a non-mainstream cluster as a word usage according to a predetermined extraction criterion for each word set group having the same basic word in the document from the classification result performed by the word classification unit 40. Extraction is performed, and a combination of a peripheral word and a base word belonging to a non-mainstream cluster is extracted as an ambiguous point that is likely to be misinterpreted such as syntax analysis. Here, the “extraction criterion” may be a criterion for classifying non-mainstream clusters as word usage. For example, the extraction criterion may be a method of extracting by a numerical criterion such as a non-mainstream cluster is a cluster in which the number of peripheral words belonging is less than an arbitrary threshold. Alternatively, as a criterion for extraction, a method of extracting on the basis of a semantic criterion such as extracting a cluster having the largest number of neighboring words and a cluster having the longest distance between clusters or an arbitrary threshold is suitable. When peripheral words are classified by fuzzy clustering, the number of belongings is handled as the sum of the degrees of belonging.

  The ambiguous point output unit 60 outputs the ambiguous points extracted by the ambiguous point extracting unit 50 for each word set group having the same basic word in the document. Here, a suitable output form is a form in which the entire document is output by clearly indicating the ambiguous points in the document by color coding or bold emphasis. In addition, the output form may be a form in which a graph in which the base word and the peripheral word are connected by a link is displayed, and the peripheral word and the link that are set as an ambiguous point are color-coded. Alternatively, as an output form, a graph that connects the base word and a keyword representing a cluster of neighboring words as a representative node with a link, and further connects each neighboring word included in the cluster as a subordinate node with the representative node is displayed, and is ambiguous. A form such as color-coding peripheral words and links below the pointed cluster may be used. In addition, as an output form, quantitative ambiguity is added to the ambiguity point based on the number of attributions of neighboring words used when extracting non-mainstream clusters, the attribution rate derived therefrom, or the distance between clusters, etc. The display may be limited only to ambiguous points that are larger than the arbitrarily set threshold. Alternatively, as an output form, color coding, emphasis by bold characters, or the size of character of a word of a graph may be given depending on the ambiguity of the ambiguity point.

  Next, the overall operation of the requested document analysis system 100 according to the first embodiment of the present invention will be described in detail with reference to the sequence diagrams of FIGS. 1 and 2.

  The word set extraction unit 10 receives input of a target document, applies morphological analysis and syntax analysis to each sentence of the document, and extracts a word set based on a predetermined extraction rule (step A1 in FIG. 2).

  For each word set, the base word extraction unit 20 extracts a base word from each word included in the word set based on a predetermined base setting rule, and the words in the word set are the base word and other words that are other words. (Step A2).

  The word similarity database 110 collects and accumulates semantic classification of words and semantic information such as synonyms and synonyms, and searches and responds to semantic information related to the meaning of the word in response to an inquiry about a specific word. (Step A3).

  The word similarity calculation unit 30 collects word sets having the same basic word from all word sets, inquires the word similarity database 110 for semantic information for each peripheral word in the word set group, and responds. Based on the semantic information thus obtained, similarity between neighboring words is calculated (step A4).

  The word classification unit 40 classifies the peripheral words by clustering the peripheral words based on the similarity between the peripheral words for each word set group having the same base word in the document (step A5).

  The ambiguous point extraction unit 50 determines, as a non-mainstream cluster, a non-mainstream cluster as a word usage according to a predetermined extraction criterion for each word set group having the same basic word in the document from the classification result performed by the word classification unit 40. Then, a combination of a peripheral word and a base word belonging to a non-mainstream cluster is extracted as an ambiguous point where a possibility of a mistake such as a dependency error is assumed (step A6).

  The ambiguous point output unit 60 outputs the ambiguous points extracted by the ambiguous point extracting unit 50 for each word set group having the same basic word in the document (step A7).

  Next, the effect of the requested document analysis system 100 according to the first embodiment of the present invention will be described.

  In the first embodiment, since the ambiguity point is extracted based on the non-mainstream cluster determined by the similarity of the usage of the words in the document or document group, the target document or document It is possible to find usage points that are considered unnatural with respect to usage specific to a group, and it is possible to extract ambiguous points such as request documents that have many unique usages.

  The required document analysis system 100 according to the first embodiment of the present invention can be realized as a required document analysis method. The requested document analysis system 100 according to the first embodiment of the present invention may be executed by a computer using a requested document analysis program.

[Embodiment 2]
Next, a second embodiment of the present invention will be described in detail with reference to the drawings.

  FIG. 3 is a block diagram showing the configuration of a requested document analysis system 100A according to the third embodiment of the present invention.

  Referring to FIG. 3, the requested document analysis system 100A according to the second exemplary embodiment of the present invention further includes a basic word versatility calculation unit 25, and the operation of the ambiguous point extraction unit is different as will be described later. Except for this, it has the same configuration as the requested document analysis system 100 according to the first embodiment shown in FIG. 1 and operates. Therefore, the reference point of 50A is given to the ambiguous point extraction unit.

  When the requested document analysis system 100A shown in the figure is realized by the above-described computer, the data processing apparatus includes a word set extraction unit 10, a basic word extraction unit 20, a basic word versatility calculation unit 25, a word similarity calculation unit 30, and a word classification. The auxiliary storage device operates as the word similarity database 110, and the output device functions as the ambiguous point output unit 60.

  The basic word versatility calculation unit 25 calculates the generality of the basic word, and the ambiguous point extraction unit 50A converts the extraction standard of the non-mainstream cluster of the neighboring word group based on the generality.

  Next, the operation of each component constituting the requested document analysis system 100A will be described.

  The basic word versatility calculating unit 25 calculates a general degree of versatility representing the generality of the basic word extracted by the basic word extracting unit 20 from the example of the basic word such as the frequency of use in a general document. Here, the “general degree” is an index that reflects whether the basic word is a general word that can be easily combined with various other words or a special word that can only be combined with a specific word. good. As a method of calculating the versatility, the number of types of vocabulary and the number of example sentences in a general dictionary, the appearance frequency in a general document group existing in books, the Web, etc., and any one or a combination thereof are calculated. Is appropriate.

  The ambiguous point extraction unit 50A calculates, for each word set group having the same basic word in the document, a threshold value set when classifying the peripheral word group into the mainstream cluster and the non-mainstream cluster based on the generality of the basic word. However, it differs from the ambiguous point extraction unit 50 shown in FIG. 1 in that an extraction criterion based on this threshold is used. Here, the “threshold value” may be set so as to change the likelihood of the number of types of peripheral words combined with the base word depending on the versatility of the base word. For example, since various semantic combinations can be assumed for neighboring word groups combined with a general-purpose key word, the threshold value considered non-mainstream is increased, and peripheral words combined with a low-level key word Since it is difficult to assume various semantic combinations for the group, the threshold value is set so as to lower the threshold value regarded as non-mainstream. The threshold value may be calculated by arbitrarily setting a function that is in a monotonically increasing relationship with the versatility of the basic word. Alternatively, the threshold value may be estimated from an empirical relationship between the versatility of the basic word and an appropriate intercluster distance.

  Other configurations of the word set extraction unit 10, the basic word extraction unit 20, the word similarity calculation unit 30, the word similarity database 110, the word classification unit 40, and the ambiguous point output unit 60 are illustrated in FIG. Since they are the same as those of the first embodiment shown in FIG.

  Next, the entire operation of the requested document analysis system 100A according to the second exemplary embodiment of the present invention will be described in detail with reference to the sequence diagrams of FIGS.

  Compared to the operation of the first embodiment described above, the operation of the second embodiment is different in that the following operation is added.

  That is, the basic word versatility calculating unit 25 calculates a general degree of versatility for the basic word extracted by the basic word extracting unit 20 that represents generality from the viewpoint of the basic word such as the frequency of use in a general document (step B1). .

  Further, the ambiguous point extraction unit 50A sets, for each word set group having the same basic word in the document, a threshold value set when classifying the peripheral word group into the mainstream cluster and the non-mainstream cluster based on the generality of the basic word. Calculate (step B2).

  The operation of other steps is the same as the operation in the first embodiment described above, and a description thereof will be omitted.

  Next, effects of the second exemplary embodiment of the present invention will be described.

  In the second embodiment, since the versatility of the base word is calculated and the likelihood of the number of types of peripheral words combined with the base word is changed due to the generality, the feature of the base word is further improved. It is possible to find usage points that are regarded as unnatural with respect to usages specific to a target document or document group based on criteria that match the criteria, and to extract ambiguous points such as required documents that have many unique usages.

  The required document analysis system 100A according to the second embodiment of the present invention can be realized as a required document analysis method. The requested document analysis system 100A according to the first embodiment of the present invention may be executed by a computer using a requested document analysis program.

  Next, with reference to FIG. 5, the operation of the requested document analysis system 100 according to the first embodiment of the present invention will be described using a specific first example.

  The purpose of the first embodiment is as follows.

  First, the requested document analysis system 100 extracts an ambiguous point A in the requested document D such as a proposal request document issued by a customer or a proposal provided by a developer in an upstream process of information system construction. The requested document analysis system 100 uses information for correction or the like as long as it is a provided document in order to read the document correctly so that there is no error in the interpretation or to confirm the correct interpretation with the document writer. Visualization enables accurate request acquisition and requirement definition.

  In the first embodiment, the requested document analysis system 100 is configured by a requested document analysis system Y and an Internet server Z as shown in FIG.

  The requested document analysis system Y operates on the PC terminal possessed by the analyst B, and through the input unit and the output unit, the input of the sentences constituting the group of documents that the analyst B wants to extract the ambiguous points, and the ambiguous Realize point A.

  The Internet server Z is connected via a communication network to a PC terminal owned by an analyst B who implements the requested document analysis system Y. The Internet server Z is a device that enables retrieval of information on semantic similarity in response to an inquiry about the similarity between words from the request document analysis system Y.

  The correspondence between FIG. 5 and FIG. 1 will be described.

  The word set extraction unit 10, the base word extraction unit 20, the word similarity calculation unit 30, the word classification unit 40, and the ambiguous point extraction unit 50 are included in the requested document analysis system Y. The ambiguous point output unit 60 operates as an output unit of the PC terminal. The word similarity database 110 is included in the Internet server Z.

  The requested document analysis system Y and the Internet server Z provided with such means operate as follows.

  The requested document analysis system Y receives an input of the document D from which the analysis operator B wants to extract the ambiguous point from the input unit. Then, the requested document analysis system Y applies morphological analysis and syntax analysis to every sentence i of the n documents i (i = 1, 2,..., N) constituting the document D, and the sentence i Are extracted, and the words corresponding to the subject Si and the predicate Vi are extracted, and a combination of the corresponding subject Si and the predicate Vi is defined as a word set Ti. Extract.

  When there are a plurality of types of subjects Si and predicates Vi in a sentence i such that the subject Sia relates to the predicate Via and the subject Sib relates to the predicate Vib, such as a heavy sentence, the word set Ti is the word set Tia {subject Sia, predicate Via. }, Tib {subject Sib, predicate Vib}, respectively. Also, in the case where the subject Sic is related to the predicate Vic and the predicate Vid, such as a compound sentence, the word set Ti is extracted separately like the word set Tic {subject Sic, predicate Vic}, Tid {subject Sic, predicate Vid}. Is done.

  Further, the requested document analysis system Y adds up the word set Ta, which is the same subject Sa, with respect to all the word sets Ti included in the document D by using the subject Si as a base, thereby obtaining the same subject Sa and subject Sa. A word set group Ga {Sa, Vaj} in which m predicates Vaj (j == 1, 2,..., M) combined are extracted. Then, the requested document analysis system Y makes an inquiry to the Internet server Z about the semantic classification of each predicate Vaj of the word set group Ga, so that the thesaurus classification system stored in the Internet server Z is used. , The semantic classification Caj to which each predicate Vaj belongs, the synonym Weaj, and the synonym Wsaj are extracted.

  The Internet server Z classifies the words according to the upper / lower relationship, the partial / whole relationship, the synonym relationship, the synonym relationship, etc. of the word, and accumulates systematic thesaurus information. In addition, the Internet server Z also provides a function such as a search engine that extracts information of an arbitrary word, so that the classification of the query target word and the synonyms and synonyms in response to the inquiry from the request document analysis system Y Are extracted and presented.

  Further, the requested document analysis system Y calculates the similarity Fa between the predicates Vaj based on the semantic classification Caj to which each predicate Vaj belongs, the synonym Weaj, and the synonym Wsaj.

  As an example of a method for calculating the similarity Fapq between the predicate Vap (j = p) and the predicate Vaq (j = q), the classification hierarchy in which the semantic classification Cap and Caq are the same classification and the rough classification hierarchy in the classification system There is a method of quantifying by the hierarchy difference. For example, in a thesaurus having a classification system consisting of four levels of major classification (first layer), middle classification (second layer), minor classification (third layer), and fine classification (fourth layer), semantic classification Cap and Caq are If they match in the small classification, “2” is used as the similarity index as the difference between the first and third layers.

  Further, as an example of calculating the similarity Fapq between the predicate Vap and the predicate Vaq more quantitatively, based on the synonyms Weap and Weaq and the synonyms Wsap and Wsaq, the synonym duplication rate Repq and the synonym duplication rate Rspq are calculated. There is a method in which the synonym weighting coefficient αe and the synonym weighting coefficient αs are set arbitrarily (αe + αs = 1, αe ≧ αs) and quantified by the weighted sum (αe × Repq + αs × Rspq). For example, in FIG. 6, the word set group Ga {Sa = “control system”, Va = “perform”, “implement”, “operation”, “use”, “development”, “construction”, “correction”, “ A part of the calculated similarity Fa is shown as “addition”, “heating”, “...”, “.

  Further, the requested document analysis system Y classifies each predicate Vaj by performing clustering that groups similar predicates Vaj based on the similarity Fa between the predicates Vaj for each word set group Ga {Sa, Vaj}. The dendrogram Ha is created.

  In the dendrogram Ha, each predicate Vaj is regarded as an initial cluster, the reciprocal of the similarity Fa is defined as the distance between each cluster (predicate Vaj), the clusters having the closest intercluster distance are defined as new clusters, and all new clusters It is created by repeating the process of combining the nearest two and creating a new cluster until all the clusters are combined into one cluster. FIG. 7 shows an example of a dendrogram Ha obtained by clustering using a dendrogram based on the similarity Fa of FIG.

  Further, for each word set group Ga {Sa, Vaj}, the requested document analysis system Y is classified based on the distance between the clusters in the dendrogram Ha, and the distance between the clusters of the predicate Vaj is arbitrarily set to a threshold N. The predicate Vax belonging to the non-mainstream cluster is extracted with the side having a larger number of predicates belonging to the cluster as the mainstream cluster and the side having the smaller number of predicates belonging to the non-mainstream cluster, and the combination with the subject Sa is defined as an ambiguous point Aa {Sa , Vax}.

  In the example of FIG. 7, when the threshold of the distance between the clusters is 5 or more, as shown in FIG. 8, the cluster is classified into a cluster to which “heating” belongs and a cluster to which all other predicates belong, and “heating” with few attributed predicates. ”Belongs to a non-mainstream cluster, Vax becomes“ heating ”, and the ambiguous point Aa {Sa =“ control system ”, Vax =“ heating ”}.

  Further, the requested document analysis system Y processes the ambiguous points Aa {Sa, Vax} corresponding to the requested document D for the ambiguous points Aa {Sa, Vax} extracted for each subject Sa by color coding or emphasis by bold type, The processed request document D is output from the output unit. In addition, when there is a predicate Vam included in the mainstream cluster in a sentence including the ambiguous point Aa {Sa, Vax}, the predicate Vam may be subjected to processing such as color coding or bold emphasis as another dependency candidate. .

  Next, the operation of the requested document analysis system 100A according to the second exemplary embodiment of the present invention will be described with reference to FIG. 9 using a specific second example.

  In the second embodiment, the requested document analysis system 100A is configured by a requested document analysis system Ya and an Internet server Z as shown in FIG.

  The requested document analysis system Ya operates on the PC terminal possessed by the analyst B, and through the input unit and the output unit, the input of the sentences constituting the group of documents that the analyst B wants to extract the ambiguous points, and the ambiguous Realize point A.

  The Internet server Z is connected via a communication network to a PC terminal owned by an analyst B who implements the requested document analysis system Ya. The Internet server Z is a device that enables retrieval of information related to similarity semantically in response to a query for similarity between words from the request document analysis system Ya.

  In the second embodiment, in addition to the operation of the first embodiment, the requested document analysis system Ya includes a basic word versatility calculating unit 25.

  That is, the correspondence between FIG. 9 and FIG. 3 is as follows.

  The word set extraction unit 10, the basic word extraction unit 20, the basic word versatility calculation unit 25, the word similarity calculation unit 30, the word classification unit 40, and the ambiguous point extraction unit 50 are included in the request document analysis system Ya Contained within. The ambiguous point output unit 60 operates as an output unit of the PC terminal. The word similarity database 110 is included in the Internet server Z.

  The requested document analysis system Ya including such a configuration adds the following operation to the first embodiment described above.

  The requested document analysis system Ya performs a Web search for each subject Si as a base word, and extracts the appearance frequency of each subject Si in a document group published on the Web. Further, the requested document analysis system Ya calculates the number of vocabularies of each subject Si from the Japanese dictionary, and calculates a value obtained by multiplying them as the versatility Ji representing the versatility of each subject Si.

  Further, the requested document analysis system Ya sets a threshold Na to be set when classifying the predicate Vaj into mainstream clusters and non-mainstream clusters based on the generality Ja of the subject Sa for each word set group Ga {Sa, Vaj}. calculate.

  As a method for calculating the threshold Na, the suitability values of the inter-cluster distances used for the classification of the non-mainstream clusters of the generality J and the predicate V calculated for various subjects S from past cases are tabulated, and the tabulated results are shown in FIG. As shown in Fig. 10, the general-purpose degree J of the word is mapped on the horizontal axis and the appropriate inter-cluster distance is plotted on the vertical axis, and the regression analysis is performed by using the general-purpose degree J of the word as an explanatory variable and the appropriate inter-cluster distance as an objective variable. A method in which an appropriate inter-cluster distance obtained by substituting the general-purpose degree Ja into the obtained regression equation is set as the threshold value Na is appropriate.

  Below, the aspect of this invention is demonstrated.

  A requested document analysis system according to a first aspect of the present invention is a system for analyzing a requested document, which applies morphological analysis and syntax analysis to each sentence of a target document, and sets a word set based on a predetermined extraction rule. A word set extraction unit for extracting the word word; for each word set, based on a predetermined rule setting rule, the word word is extracted from each word included in the word set, and the word set is the word word and the surrounding words that are other words A key word extraction unit that categorizes the words into words; accumulates semantic information of words and synonyms, synonyms, etc., and searches for semantic information related to the meaning and usage of a word in response to a query about a specific word A word similarity database that responds; a group of word sets having the same basic word from all word sets, and the above word for each peripheral word in the word set group A word similarity calculation unit that inquires semantic information in the similarity database and calculates similarity between words based on the returned semantic information; between adjacent words for each word set group having the same basic word in the document A word classification unit that classifies peripheral words based on similarity of words and classifies the peripheral words; for each word set group that has the same basic word in the document, it is extracted from the classification results performed by this word classification unit A non-mainstream cluster is extracted as a non-mainstream cluster as a word usage according to the criteria, and an ambiguous point extraction unit that extracts a combination of a peripheral word and a base word belonging to the non-mainstream cluster as an ambiguous point that is assumed to have a potential error. And an ambiguous point output unit that outputs an ambiguous point extracted for each word set group having the same basic word in the document. ; Comprises.

  The requested document analysis system according to the first aspect of the present invention calculates a versatility representing the versatility of the basic word extracted by the basic word extraction unit, such as the frequency of use in a general document, as seen from the example of the basic word. You may further provide a basic word general-purpose degree calculation part. In this case, the ambiguous point extraction unit is set when classifying peripheral word groups into mainstream clusters and non-mainstream clusters for each word set group having the same basic word in the document based on the generality of the basic word. A threshold value to be calculated may be calculated, and an extraction criterion based on this threshold value may be used.

  In the required document analysis system, the basic word versatility calculating unit includes the number of vocabulary types and the number of example sentences in a general dictionary, the appearance frequency in a general document group existing in a book, the Web, and any of these. The general degree calculated by one or a combination may be used. In this case, the ambiguous point extraction unit uses a regression equation based on the versatility of the basic word and the actual results of the distance between the clusters as the threshold to be set when the peripheral word group is classified into the mainstream cluster and the non-mainstream cluster. It is preferable to calculate from

  In the required document analysis system, the basic word extraction unit is configured to determine the basic word and the peripheral based on a basic setting rule in which the related word or subject is the basic word, the word or predicate involved, the particle that connects the subject and the predicate, or the like. Words may be extracted. The word set extraction unit may extract a combination of words having a dependency relationship as a word set. The word similarity database may be a thesaurus in which words are classified and organized according to the upper / lower relationship, partial / whole relationship, synonym relationship, synonym relationship, and the like. The word similarity calculation unit extracts the types and number of co-occurrence words that co-occur with each neighboring word in a sentence on a general Web or document, and calculates the overlapping rate of the co-occurrence words between the neighboring words. It may be used as an indicator. The ambiguous point extraction unit may extract a cluster having the largest number of neighboring words and a cluster having the longest distance between the clusters or farther than an arbitrary threshold.

  In the requested document analysis system, the ambiguous point output unit connects a keyword representing a cluster of neighboring words with a link as a representative node, and further connects each neighboring word included in the cluster with a representative node as a subordinate node. The surrounding words and links below the cluster displayed as the ambiguous point may be color-coded. Instead, the ambiguous point output unit adds quantitative ambiguity to the ambiguous points based on the number of surrounding words used when extracting non-mainstream clusters, the attribution rate derived from them, or the distance between clusters. However, the display may be limited to only the ambiguous points whose ambiguity is larger than the arbitrarily set threshold, or depending on the ambiguity of the ambiguous points, color coding, bold emphasis, or the size of the words of the graph may be given strength .

  A request document analysis method according to a second aspect of the present invention is a method for analyzing a request document, wherein morpheme analysis and syntax analysis are applied to each sentence of a target document, and a word set is based on a predetermined extraction rule. A word set extraction step for extracting the word set; for each word set, based on a predetermined basic axis setting rule, a basic word is extracted from each word included in the word set, and the word set is a basic word and other words that are other words A word extraction step that classifies the words into a word similarity database that accumulates semantic classification of words and semantic information such as synonyms and synonyms, and is related to the meaning and usage of words for inquiries about specific words Searching for semantic information and responding; collecting word sets having the same basic word from all word sets, and collecting each of the surrounding words of the word set A word similarity calculation step of querying the word similarity database for semantic information and calculating similarity between words based on the returned semantic information; a set of words having the same basic word in the document For each word, a word classification step of classifying the peripheral words based on the similarity between the peripheral words and classifying the peripheral words; a word having the same base word in the document from the classification result performed in this word classification step For each set group, a non-mainstream cluster is extracted as a non-mainstream cluster as a word usage according to the extraction criteria, and a combination of a peripheral word and a base word belonging to the non-mainstream cluster is assumed as an ambiguous point that may be a misunderstanding point. A fuzzy point extraction step to extract; a word set having the same basic word in the document in this fuzzy point extraction step Including; a fuzzy point output step of outputting the ambiguous points extracted for each.

  In the request document analysis method according to the second aspect of the present invention, the versatility indicating the versatility of the basic word extracted in the basic word extraction step, such as the frequency of use in a general document, as seen from the example of the basic word is calculated. A basic word versatility calculation step may be further included. In this case, the ambiguous point extraction step is set when classifying the peripheral word group into the mainstream cluster and the non-mainstream cluster based on the versatility of the base word for each word set group having the same base word in the document. A threshold value to be calculated may be calculated, and an extraction criterion based on this threshold value may be used.

  In the required document analysis method, the basic word versatility calculation step includes the number of types of vocabulary and the number of example sentences in a general dictionary, the appearance frequency in a general document group existing in a book or the Web, and any of these The general degree calculated by one or a combination may be used. In this case, the ambiguous point extraction step uses a regression equation based on the versatility of the basic word and the actual results of the distance between the clusters, as the threshold value set when classifying the peripheral word group into the mainstream cluster and the non-mainstream cluster. It is preferable to calculate from

  In the required document analysis method, the basic word extraction step includes the basic word and the subject based on the basic setting rule in which the related word or subject is a basic word, the related word or predicate, the particle that connects the subject and the predicate, and the like. Peripheral words may be extracted. In the word set extraction step, a combination of words having a dependency relationship may be extracted as a word set. The word similarity database may be a thesaurus in which words are classified and organized according to the upper / lower relationship, partial / whole relationship, synonym relationship, synonym relationship, and the like. The word similarity calculation step extracts the types and number of co-occurrence words that co-occur with each neighboring word in a sentence on a general Web or document, and calculates the overlapping rate of the co-occurrence words between the neighboring words. It may be used as an indicator. The fuzzy point extracting step may extract a cluster having the largest number of neighboring words and a cluster having the longest distance between the clusters or farther than an arbitrary threshold.

  In the requested document analysis method, the ambiguous point output step includes a graph in which a keyword representing a cluster of neighboring words is linked with a representative node as a representative node, and each neighboring word included in the cluster is linked with a representative node as a dependent node. The surrounding words and links below the cluster displayed as the ambiguous point may be color-coded. Instead, the ambiguity point output step adds quantitative ambiguity to the ambiguity point based on the number of surrounding words used to extract the non-mainstream cluster, the affiliation rate derived from it, or the distance between clusters. However, the display may be limited to only the ambiguous points whose ambiguity is larger than the arbitrarily set threshold, or depending on the ambiguity of the ambiguous points, color coding, bold emphasis, or the size of the words of the graph may be given strength .

  The requirement document analysis program according to the third aspect of the present invention is a requirement document analysis program that causes a computer to analyze a requirement document, and applies morphological analysis and syntax analysis to each sentence of the target document to the computer. A word set extraction procedure for extracting a word set based on a predetermined extraction rule; for each word set, a basic word is extracted from each word included in the word set based on a predetermined basic setting rule; Base word extraction procedure for classifying words into other words and neighboring words; word semantic database that stores semantic classification of words and semantic information such as synonyms and synonyms; On the other hand, a procedure for retrieving and responding to semantic information related to the meaning and usage of a word; a word having the same basic word from all word sets Word similarity calculation that collects word groups, queries the word similarity database for semantic information for each peripheral word in the word set group, and calculates similarity between words based on the responded semantic information Procedure: A word classification procedure for classifying neighboring words by clustering neighboring words based on the similarity between neighboring words for each word set group having the same basic word in the document; implemented in this word classification step From the classification results, for each word set group that has the same basic word in the document, non-mainstream clusters are extracted as non-mainstream clusters as word usage according to the extraction criteria, and combinations of peripheral words and basic words belonging to non-mainstream clusters A fuzzy point extraction procedure for extracting fuzzy points as possible fuzzy points that are likely to have a dependency error; A fuzzy point output procedure to output the ambiguous points extracted for each word set group with the same base shaft words in writing; to execution.

  The requested document analysis program according to the third aspect of the present invention is a general-purpose computer that represents the versatility of the basic word extracted in the basic word extraction procedure in terms of the use of the basic word such as the frequency of use in a general document. The basic word general-purpose degree calculation procedure for calculating the degree may be further executed. In this case, the ambiguous point extraction procedure is set when classifying peripheral word groups into mainstream clusters and non-mainstream clusters based on the generality of the base word for each word set group having the same base word in the document. A threshold value to be calculated may be calculated, and an extraction criterion based on this threshold value may be used.

  In the required document analysis program, the basic word versatility calculation procedure includes the number of types of vocabulary and the number of example sentences in a general dictionary, the appearance frequency in a general document group existing in a book or the Web, and any of these. The general degree calculated by one or a combination may be used. In this case, the ambiguous point extraction procedure uses a regression formula based on the versatility of the basic word and the actual results of the distance between the clusters as the threshold value set when classifying the peripheral word group into the mainstream cluster and the non-mainstream cluster. It is preferable to calculate from

  In the requirement document analysis program, the basic word extraction procedure is based on the basic word setting rule in which the word or subject is the basic word, the word or predicate to be engaged, the particle connecting the subject and the predicate, and the like as the peripheral word. Peripheral words may be extracted. In the word set extraction procedure, a combination of words having a dependency relationship may be extracted as a word set. The word similarity database may be a thesaurus in which words are classified and organized according to the upper / lower relationship, partial / whole relationship, synonym relationship, synonym relationship, and the like. The above word similarity calculation procedure extracts the types and number of co-occurrence words that co-occur with each neighboring word in a sentence on a general Web or document, and calculates the overlapping rate of co-occurrence words between neighboring words. It may be used as an indicator. The ambiguous point extraction procedure may extract a cluster having the largest number of neighboring words and a cluster having the longest distance between the clusters or farther than an arbitrary threshold.

  In the requested document analysis program, the ambiguous point output procedure includes a graph in which keywords representing a cluster of neighboring words are linked by a representative node, and each neighboring word included in the cluster is linked to a representative node as a subordinate node. Peripheral words and links below the cluster displayed as ambiguous points may be color-coded. Instead, the above ambiguous point output procedure adds quantitative ambiguity to the ambiguous point based on the number of neighboring words used to extract the non-mainstream cluster, the attribution rate derived from it, or the distance between clusters. However, the display may be limited to only the ambiguous points whose ambiguity is larger than the arbitrarily set threshold, or depending on the ambiguity of the ambiguous points, color coding, bold emphasis, or the size of the words of the graph may be given strength .

  The present invention has been described above with reference to the embodiments (and examples), but the present invention is not limited to the above embodiments (and examples). Various changes that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention.

  According to the present invention, it is possible to support understanding, creation, and correction of documents by excluding ambiguity of documents for various documents exchanged in work such as requirement definition in software and system development. It can be applied to applications related to streamlining system development, such as reducing rework and improving customer satisfaction.

DESCRIPTION OF SYMBOLS 10 Word set extraction part 20 Basic word extraction part 25 Basic word versatility calculation part 30 Word similarity calculation part 40 Word classification part 50, 50A Ambiguous point extraction part 60 Ambiguous point output part 100, 100A Request document analysis system 110 Word similarity Database D Request document A Ambiguous point Y, Ya Request document analysis system Z Internet server

Claims (30)

A system for analyzing requirements documents,
A word set extraction unit that applies morphological analysis and syntax analysis to each sentence of a target document and extracts a word set based on a predetermined extraction rule;
A base word that extracts a base word from each word included in the word set and classifies the word set into a base word and peripheral words that are other words based on a predetermined base setting rule for each word set An extractor;
A word similarity database that accumulates semantic information of words, synonyms, synonyms, etc., retrieves and responds to inquiries about specific words, and retrieves semantic information related to the meaning and usage of words;
Summarizing word sets having the same basic word from all word sets, inquiring semantic information from the word similarity database for each of the surrounding words of the word set group, based on the responded semantic information, A word similarity calculator for calculating similarity between words;
For each word set group having the same basic word in the document, a word classification unit that performs clustering of the peripheral words based on the similarity between the peripheral words and classifies the peripheral words,
For each word set group having the same basic word in the document, a non-mainstream cluster is extracted as a non-mainstream cluster as a word usage according to the extraction criteria from the classification result performed in the word classification unit, and the peripherals belonging to the non-mainstream cluster An ambiguous point extraction unit that extracts a combination of a word and a basic word as an ambiguous point that is assumed to have a possibility of a mistake, etc .;
An ambiguous point output unit that outputs an ambiguous point extracted for each word set group having the same basic word in the document in the ambiguous point extracting unit;
A requirement document analysis system characterized by comprising: For the base word extracted by the base word extraction unit, the base word general degree calculation unit further calculates a general degree representing generality as seen from the use example of the base word such as the frequency of use in a general document,
For each word set group having the same basic word in the document, the ambiguous point extracting unit sets a threshold value to be set when classifying the peripheral word group into the mainstream cluster and the non-mainstream cluster based on the generality of the basic word. The request document analysis system according to claim 1, wherein an extraction criterion based on the threshold is calculated and used. The basic word general-purpose degree calculation unit calculates the number of vocabulary types and the number of example sentences in a general dictionary, the appearance frequency in a general document group existing in a book or the Web, and any one or a combination thereof. Use the versatility to be
The ambiguous point extraction unit calculates a threshold value that is set when the neighboring word group is classified into the mainstream cluster and the non-mainstream cluster from a regression equation based on the versatility of the basic word and the actual results of the distance between the clusters. The requirement document analysis system according to claim 2, wherein:   The basic word extraction unit extracts the basic word and the peripheral word based on the basic setting rule in which the related word or subject is the basic word, the word or predicate involved, the particle that connects the subject and the predicate, and the like as the peripheral word; The requirement document analysis system according to any one of claims 1 to 3, wherein:   The request document analysis system according to claim 1, wherein the word set extraction unit extracts a combination of words having a dependency relationship as a word set.   6. The thesaurus according to claim 1, wherein the word similarity database is a thesaurus in which words are classified and organized by upper / lower relationship, partial / whole relationship, synonym relationship, synonym relationship, etc. of words. The requirement document analysis system according to any one of the above.   The word similarity calculation unit extracts the types and the number of co-occurrence words that co-occur with each neighboring word in a sentence on a general Web or document, and calculates the overlapping rate of the co-occurrence words between the neighboring words. The request document analysis system according to claim 1, wherein the request document analysis system is used as an index.   8. The ambiguous point extracting unit extracts a cluster having the largest number of neighboring words and a cluster having the longest distance between clusters or a cluster farther than an arbitrary threshold. Requirement document analysis system described in.   The ambiguous point output unit displays a graph in which a keyword representing a cluster of neighboring words is linked by a link as a representative node, and each neighboring word included in the cluster is linked to a representative node as a subordinate node, and the ambiguous point and The request document analysis system according to claim 1, wherein a peripheral word and a link below the determined cluster are color-coded.   The ambiguous point output unit adds a quantitative ambiguity to the ambiguous point based on the number of surrounding words used when extracting the non-mainstream cluster, the degree of attribution derived therefrom, or the distance between clusters. The display is limited to vague points that are larger than a threshold value that is arbitrarily set, or the ambiguity of the fuzzy points gives color strength, bold emphasis, or the size of the words in the graph. The request document analysis system according to any one of claims 1 to 8. A method for analyzing a requirements document, comprising:
A word set extraction step of applying morphological analysis and syntax analysis to each sentence of a target document and extracting a word set based on a predetermined extraction rule;
A base word that extracts a base word from each word included in the word set and classifies the word set into a base word and peripheral words that are other words based on a predetermined base setting rule for each word set An extraction step;
Searching and responding to semantic information related to the meaning and usage of a word in response to an inquiry about a specific word in a word similarity database that accumulates semantic information such as semantic classification and synonyms and synonyms of the word; and
Summarizing word sets having the same basic word from all word sets, inquiring semantic information from the word similarity database for each of the surrounding words of the word set group, based on the responded semantic information, A word similarity calculation step for calculating similarity between words;
For each word set group having the same basic word in the document, a word classification step for classifying the peripheral words by performing clustering of the peripheral words based on the similarity between the peripheral words,
From the classification result performed in the word classification step, for each word set group having the same basic word in the document, a non-mainstream cluster is extracted as a non-mainstream cluster as a word usage according to the extraction criterion, and the surroundings belonging to the non-mainstream cluster An ambiguous point extraction step of extracting a combination of a word and a basic word as an ambiguous point that is assumed to be a possibility of a mistake, etc .;
An ambiguous point output step for outputting the ambiguous point extracted for each word set group having the same basic word in the document in the ambiguous point extracting step;
A requirement document analysis method comprising: For the base word extracted in the base word extraction step, it further includes a base word versatility calculation step for calculating a general degree representing versatility from the viewpoint of examples of the base word such as frequency of use in general documents,
The threshold value set when the fuzzy point extraction step classifies peripheral word groups into mainstream clusters and non-mainstream clusters based on the generality of the base word for each word set group having the same base word in the document. 12. The request document analysis method according to claim 11, wherein an extraction criterion based on the threshold is calculated and used. The basic word versatility calculation step calculates the number of types of vocabulary and the number of example sentences in a general dictionary, the appearance frequency in a general document group existing in a book or the Web, and any one or a combination thereof. Use the versatility to be
The fuzzy point extracting step calculates a threshold value set when classifying the peripheral word group into the mainstream cluster and the non-mainstream cluster from a regression equation based on the versatility of the basic word and the actual results of the distance between the clusters. The request document analysis method according to claim 12, wherein:   The basic word extraction step extracts the basic word and the peripheral word based on the basic setting rule in which the word or subject is the basic word, the word or predicate involved, the particle connecting the subject and the predicate, and the like. 14. The request document analysis method according to any one of claims 11 to 13, wherein:   15. The request document analysis method according to claim 11, wherein the word set extraction step extracts a combination of words having a dependency relationship as a word set.   The word similarity database is a thesaurus in which words are classified and organized by upper / lower relationship, partial / whole relationship, synonym relationship, synonym relationship, etc. of words. The required document analysis method according to any one of the above items.   The word similarity calculation step extracts the types and number of co-occurrence words that co-occur with each neighboring word in a sentence on a general Web or document, and calculates the overlapping rate of the co-occurrence words between the neighboring words. The request document analysis method according to claim 11, wherein the request document analysis method is used as an index.   18. The ambiguous point extracting step extracts a cluster having the largest number of neighboring words and a cluster having the longest distance between the clusters or an arbitrary threshold. Requirement document analysis method described in 1.   The ambiguous point output step displays a graph in which a keyword representing a cluster of neighboring words is linked as a representative node, and each neighboring word included in the cluster is linked as a subordinate node to a representative node, The request document analysis method according to any one of claims 11 to 18, characterized in that peripheral words and links below the determined cluster are color-coded.   The ambiguity point output step adds a quantitative ambiguity to the ambiguity point based on the number of surrounding words used when extracting the non-mainstream cluster, the affiliation rate derived therefrom, or the distance between clusters, etc. The display is limited to vague points that are larger than a threshold value that is arbitrarily set, or the ambiguity of the fuzzy points gives color strength, bold emphasis, or the size of the words in the graph. The request document analysis method according to any one of claims 11 to 18. A request document analysis program for causing a computer to analyze a request document, wherein the computer
A word set extraction procedure that applies morphological analysis and syntax analysis to each sentence of a target document and extracts a word set based on a predetermined extraction rule;
A base word that extracts a base word from each word included in the word set and classifies the word set into a base word and peripheral words that are other words based on a predetermined base setting rule for each word set Extraction procedure;
A procedure for retrieving and responding to semantic information related to the meaning and usage of a word in a query related to a specific word in a word similarity database that accumulates semantic information such as semantic classification and synonyms and synonyms of the word,
Summarizing word sets having the same basic word from all word sets, inquiring semantic information from the word similarity database for each of the surrounding words of the word set group, based on the responded semantic information, A word similarity calculation procedure for calculating similarity between words;
For each word set group having the same basic word in the document, a word classification procedure for performing clustering of the peripheral words based on the similarity between the peripheral words and classifying the peripheral words,
From the classification result performed in the word classification step, for each word set group having the same basic word in the document, a non-mainstream cluster is extracted as a non-mainstream cluster as a word usage according to the extraction criterion, and the surroundings belonging to the non-mainstream cluster An ambiguous point extraction procedure that extracts a combination of a word and a basic word as an ambiguous point that is assumed to be a possibility of a mistake, etc.,
An ambiguous point output procedure for outputting the ambiguous point extracted for each word set group having the same basic word in the document in the ambiguous point extracting step;
Request document analysis program to execute. Causing the computer to further execute a baseline word versatility calculation procedure for calculating a versatility representing versatility from a viewpoint word example, such as a use frequency in a general document, with respect to a baseline word extracted in the baseline word extraction procedure;
For the word set group having the same basic word in the document, the fuzzy point extracting procedure sets a threshold value to be set when classifying the peripheral word group into the mainstream cluster and the non-mainstream cluster based on the generality of the basic word. The requirement document analysis program according to claim 21, wherein an extraction criterion based on the threshold is calculated and used. The basic word versatility calculation procedure is calculated based on the number of types of vocabulary and the number of example sentences in a general dictionary, the appearance frequency in a general document group existing in a book or the Web, and any one or a combination thereof. Use the versatility to be
The fuzzy point extraction procedure calculates a threshold value that is set when the neighboring word group is classified into the mainstream cluster and the non-mainstream cluster from a regression equation based on the versatility of the basic word and the actual distance between the clusters. The request document analysis program according to claim 22, wherein:   The basic word extraction procedure is to extract the basic word and the peripheral word based on the basic setting rule in which the word or subject is the basic word, the word or predicate to be engaged, the particle that connects the subject and the predicate, and the like. The requirement document analysis program according to any one of claims 21 to 23.   25. The request document analysis program according to claim 21, wherein the word set extraction procedure extracts a combination of words having a dependency relationship as a word set.   26. The thesaurus in which the word similarity database is a thesaurus in which words are classified and organized by upper / lower relationship, partial / whole relationship, synonym relationship, synonym relationship, etc. of words. The requirement document analysis program according to any one of the above items.   The word similarity calculation procedure extracts the type and number of co-occurrence words that co-occur with each neighboring word in a sentence on a general Web or document, and calculates the overlapping rate of co-occurrence words between neighboring words. 27. The request document analysis program according to claim 21, wherein the request document analysis program is used as an index.   28. The ambiguous point extraction procedure, wherein a cluster having the largest number of neighboring words and a cluster having the longest distance between the clusters or a cluster farther than an arbitrary threshold is extracted. Requirement document analysis program described in   The ambiguous point output procedure displays a graph in which keywords representing a cluster of neighboring words are linked with a representative node as a representative node, and each neighboring word included in the cluster is linked with a representative node as a subordinate node, 29. The requested document analysis program according to claim 21, wherein the peripheral words and links below the set cluster are color-coded.   The ambiguous point output procedure adds a quantitative ambiguity to the ambiguous point based on the number of attributions of neighboring words used when extracting non-mainstream clusters, the attribution rate derived therefrom, or the distance between clusters. The display is limited to vague points that are larger than a threshold value that is arbitrarily set, or the ambiguity of the fuzzy points gives color strength, bold emphasis, or the size of the words in the graph. The request document analysis program according to any one of claims 21 to 28.

Images