JP2005293261A - Evaluation apparatus, evaluation method, and evaluation program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To realize evaluation work versatilely which is efficient and good in quality in each subcategory timely. <P>SOLUTION: A natural language processing part 201 performs syntax analyses and semantic analyses of each sentence in a carrier related document; a mapping rule storage part 202 stores mapping rules for associating evaluation items of each subcategory (a skill item, motivation item, and mental item) with question items of a question/answer table; a matching part 203 performs matching of each sentence analyzed by the natural language processing part 201 with the mapping rules; an application processing part 205 creates answers to the question items associated by the mapping rules to store them in a question/answer information storage part 206; and a mapping part 208 maps the answers stored in the question/answer information storage part 209 (a data base of each subcategory) onto data items to store them. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to an evaluation apparatus, an evaluation method, and an evaluation program for extracting and evaluating evaluation items from an input sentence described in a natural language.

  Conventionally, when conducting a survey or analysis for a specific purpose, surveys and interviews are conducted on respondents for the questions and survey items (evaluation items) prepared in advance. It was common. For example, in the case of evaluating individual skills, necessary information was collected by allowing the evaluator to answer a series of questions using a predetermined question / answer table (for example, patents). Reference 1).

  In addition, even when these survey / analysis operations are systematized, it is common practice to have respondents answer while sequentially displaying questions prepared in advance. Furthermore, there is a natural language processing technique that can be used when a necessary skill item is extracted from a document written in a natural language such as a business history instead of a question / answer table (see, for example, Patent Documents 2 and 3).

JP 2002-140451 A JP 2003-15514 A JP 2002-230344 A

  However, the conventional technology that relies on the question / answer table requires a large amount of data creation and collection in advance, so the problem is that timely evaluation and analysis cannot be performed, and new data is created. There is a problem that the cost for generating will occur. In addition, since the collected data is mostly direct response information for a specific purpose or survey, there is a problem that it is difficult to reuse.

  In this way, there are many problems with evaluations that depend on the question / answer table. Therefore, it is more difficult to collect information about the evaluation items from the evaluated person in a fixed format such as the question / answer table. It can be said that it is a desirable format for information gathering because it is possible to obtain evaluation items in a free form and to obtain many evaluation items accurately. In other words, opinions and information (e.g., posted information by consumers, viewers, readers, HR related career history, career information, etc.) freely written or posted by the evaluator (subject) Compared to the case where questions are prepared in advance, the description range is broad and a lot of information useful for evaluation and analysis is included.

  Therefore, it is considered possible to improve the efficiency and quality of the survey by using the work history, etc. freely described by the evaluator as input data for the survey and analysis work described above. The larger the amount of opinions and information by the person, the more difficult it is to manually check. Therefore, as a result, these business resumes are often not used.

  For this reason, a method for extracting necessary evaluation items from a document written in a natural language such as a business history record by applying natural language processing technology instead of a manual check work can be considered. In the application of language processing technology, only inferior quality evaluation items are extracted based on character string matching, and the evaluation items are accurately extracted with high quality by analyzing the structure and contents of the document. There is a problem that it has not been realized.

  Also, for example, documents in career-related categories (business resumes, career survey forms, personnel evaluation forms, etc.) include information on multiple subcategories such as retained skill information and motivation / application information or mental information in the same document. In many cases, the conventional system is a system that processes only one of the sub-categories such as skill evaluation, motivation evaluation, or mental evaluation.

  Accordingly, the present invention has been made to solve the above-described problems of the prior art, and is an evaluation apparatus and an evaluation device that can generally implement timely, efficient, and high-quality evaluation work in each subcategory. The object is to provide a method and evaluation program.

  In order to solve the above-described problems and achieve the object, the invention according to claim 1 is an evaluation device that extracts and evaluates evaluation items from an input sentence described in a natural language, and includes a plurality of subcategories in a certain category. Evaluation result storage means for storing the evaluation results obtained for the evaluation items of each subcategory for each, and the results of natural language analysis for the sentence including the evaluation items for each evaluation item to be extracted in each subcategory As for each subcategory, a mapping rule storage means for storing a mapping rule that defines an analysis result, an evaluation result for the evaluation item, and a storage destination for storing the evaluation result in the evaluation result storage means Stored in the mapping rule storage means from the analysis results obtained by natural language analysis of mixed input statements Mapping processing means for extracting the evaluation items of each subcategory using the mapping rule and storing the evaluation results corresponding to the extracted evaluation items in the corresponding storage destination in the evaluation result storage means. Features.

  The invention according to claim 2 is characterized in that, in the above invention, the mapping rule storage means stores a mapping rule defined by a common format among the subcategories.

  Further, in the invention according to claim 3, in the above invention, when there is an evaluation item in which the analysis result is common among a plurality of subcategories, the mapping rule storage means associates with the analysis result, and A mapping rule that defines the evaluation result and the storage destination for each of a plurality of subcategories is stored.

  The invention according to claim 4 is the conversion rule storage for storing a conversion rule for converting the evaluation result of each subcategory stored in the evaluation result storage means into a predetermined evaluation table for each subcategory in the above invention. And an evaluation table creation means for creating an evaluation table for each subcategory from the evaluation result for each subcategory stored in the evaluation result storage means, using conversion rules for each subcategory stored in the conversion rule storage means, Is further provided.

  Further, in the invention according to claim 5, in the above invention, the conversion rule storage means further stores a comprehensive evaluation rule for combining the evaluation results of a plurality of subcategories into a predetermined evaluation table, The evaluation table creating means creates a comprehensive evaluation table from the evaluation results of a plurality of subcategories stored in the evaluation result storage means, using the comprehensive evaluation rules stored in the conversion rule storage means. And

  The invention according to claim 6 is the above invention, wherein the evaluation result storage means divides into a plurality of categories and stores the evaluation results of each subcategory for each of a plurality of subcategories belonging to each category. The mapping rule storage means stores a mapping rule for each evaluation item to be extracted in each subcategory belonging to the plurality of categories, and the mapping processing means is mixed for each subcategory belonging to any category. The input sentence described above is extracted and stored using the corresponding category mapping rule.

  The invention according to claim 7 further includes a receiving unit that receives a category designation for designating which category of the input sentence according to the above invention, and the mapping processing unit stores the mapping rule in the mapping rule storage unit. Of the stored mapping rules for each category, extraction and storage are performed using only the category mapping rules relating to the category designation received by the receiving means.

  The invention according to claim 8 further includes determination means for determining which category the input sentence is in the above invention, wherein the mapping processing means stores each of the mapping rule storage means. Extraction and storage are performed using only the category mapping rule determined by the determination unit among the category mapping rules.

  Further, in the invention according to claim 9, in the above invention, the mapping rule storage means uses a syntax / semantic structure assumed as a result of syntax / semantic analysis for a sentence including the evaluation item as the analysis result. The specified mapping rule is stored.

  The invention according to claim 10 is the mapping rule according to the above invention, wherein the mapping rule storage means defines, as the analysis result, a word / phrase list that is assumed as a result of a morphological analysis for a sentence including the evaluation item. Is stored.

  The invention according to claim 11 is the above invention, wherein, for each of a plurality of subcategories, a question relating to an evaluation item of each subcategory, an evaluation result obtained as an answer to the question, and the evaluation result as the evaluation result Question / answer storage means for storing a storage destination when storing in the storage means, and the mapping rule storage means replaces the storage destination of the evaluation result storage means with the evaluation result as the question / answer storage means The mapping rule that defines the storage location when storing in the storage is stored, and the mapping processing means stores the evaluation result corresponding to the evaluation item extracted using the mapping rule in the corresponding storage destination in the question / answer storage means And storing the evaluation result in a storage location of the evaluation result storage means stored in advance in the storage location. To.

  Further, the invention according to claim 12 is the above invention, wherein a question with an evaluation result not yet stored in the question / answer storage means even after extraction of an evaluation item using the mapping rule is answered to the question. Answering complementing means for complementing the evaluation result as, and when the evaluation result is complemented by the answer complementing means, the mapping processing means stores the evaluation result storing means stored in advance in the complementing destination. The evaluation result is also stored in a storage destination.

  The invention according to claim 13 further comprises speech recognition means for recognizing speech information mixedly heard for each subcategory and creating the input sentence in the above invention, wherein the mapping processing means comprises: The input sentence created by the voice recognition means is extracted and stored.

  The invention according to claim 14 is an evaluation method for extracting and evaluating evaluation items from an input sentence described in a natural language, and obtained for each subcategory evaluation item for each of a plurality of subcategories in a certain category. An evaluation result storage step for storing an evaluation result storage means for storing the evaluation result obtained, and for each evaluation item to be extracted in each subcategory, it is assumed as a result of natural language analysis for a sentence including the evaluation item A mapping rule holding step for holding an analysis result, an evaluation result for the evaluation item, and a mapping rule storage means for storing mapping rules each defining a storage destination when the evaluation result is stored in the evaluation result storage means; Based on the analysis results obtained by natural language analysis of the input sentences described in a mixed manner for each subcategory, the above-mentioned map is used. A mapping processing step of extracting the evaluation items of each subcategory using the mapping rule stored in the mapping rule storage means, and storing the evaluation results corresponding to the extracted evaluation items in the corresponding storage destination in the evaluation result storage means; It is characterized by including.

  The invention according to claim 15 is characterized in that, in the above-mentioned invention, the mapping rule storage means stores a mapping rule defined by a common format among the subcategories.

  Further, in the invention according to claim 16, in the above invention, when there is an evaluation item in which the analysis result is common among a plurality of subcategories, the mapping rule storage unit associates the analysis result with the analysis result, and A mapping rule that defines the evaluation result and the storage destination for each of a plurality of subcategories is stored.

  The invention according to claim 17 is the conversion rule storage for storing a conversion rule for converting the evaluation result of each subcategory stored in the evaluation result storage means into a predetermined evaluation table for each subcategory in the above invention. Using the conversion rule holding step for holding means and the conversion rules for each subcategory stored in the conversion rule storage means, an evaluation table for each subcategory is created from the evaluation results for each subcategory stored in the evaluation result storage means And an evaluation table creating step.

  Further, the invention according to claim 18 is the above invention, wherein the conversion rule storage means further stores a comprehensive evaluation rule for comprehensively converting the evaluation results of a plurality of subcategories into a predetermined evaluation table, In the evaluation table creation step, a comprehensive evaluation table is created from the evaluation results of a plurality of subcategories stored in the evaluation result storage unit using the comprehensive evaluation rules stored in the conversion rule storage unit. And

  The invention according to claim 19 is the above invention, wherein the evaluation result storage means is divided into a plurality of categories and stores the evaluation results of each subcategory for each of a plurality of subcategories belonging to each category. The mapping rule storage means stores a mapping rule for each evaluation item to be extracted in each subcategory belonging to the plurality of categories, and the mapping processing step is mixed for each subcategory belonging to any category. The input sentence described above is extracted and stored using the corresponding category mapping rule.

  The invention according to claim 20 further includes an accepting step of accepting category designation for designating which category the input sentence relates to in the above invention, wherein the mapping processing step is stored in the mapping rule storage means. Of the stored mapping rules for each category, extraction and storage are performed using only the category mapping rules related to the category designation received in the receiving step.

  The invention according to claim 21 further includes a determination step of determining which category the input sentence is in the above invention, wherein the mapping processing step is stored in each of the mapping rule storage means. Extraction and storage are performed using only the category mapping rules determined in the determination step among the category mapping rules.

  The invention according to claim 22 is an evaluation program for causing a computer to execute an evaluation method for extracting and evaluating evaluation items from an input sentence described in a natural language, and each subcategory for each of a plurality of subcategories in a certain category. An evaluation result storage procedure for storing an evaluation result storage means for storing evaluation results obtained for the evaluation item, and a natural language for a sentence including the evaluation item for each evaluation item to be extracted in each subcategory Holding a mapping rule storage means for storing a mapping rule that defines an analysis result assumed as a result of analysis, an evaluation result for the evaluation item, and a storage destination for storing the evaluation result in the evaluation result storage means The mapping rule retention procedure to be used and the input sentence mixed for each subcategory From the analysis result obtained by the analysis, the evaluation item of each subcategory is extracted using the mapping rule stored in the mapping rule storage unit, and the evaluation result corresponding to the extracted evaluation item is stored in the evaluation result storage unit. A mapping processing procedure stored in a corresponding storage destination is executed by a computer.

  The invention according to claim 23 is characterized in that, in the above-mentioned invention, the mapping rule storage means stores a mapping rule defined by a common format among the subcategories.

  In the invention according to claim 24, in the above invention, when there is an evaluation item in which the analysis result is common among a plurality of subcategories, the mapping rule storage unit associates the analysis result with the analysis result, and A mapping rule that defines the evaluation result and the storage destination for each of a plurality of subcategories is stored.

  In the invention according to claim 25, in the above invention, a conversion rule storage for storing a conversion rule for converting an evaluation result of each subcategory stored in the evaluation result storage means into a predetermined evaluation table for each subcategory. An evaluation table for each subcategory is created from the evaluation results for each subcategory stored in the evaluation result storage means, using the conversion rule holding procedure for holding means and the conversion rules for each subcategory stored in the conversion rule storage means And an evaluation table creation procedure to be executed by a computer.

  In the invention according to claim 26, in the above invention, the conversion rule storage unit further stores a general evaluation rule for integrating the evaluation results of a plurality of subcategories into a predetermined evaluation table, In the evaluation table creation procedure, a comprehensive evaluation table is created from the evaluation results of a plurality of subcategories stored in the evaluation result storage unit using the comprehensive evaluation rules stored in the conversion rule storage unit. And

  According to a twenty-seventh aspect of the present invention, in the above invention, the evaluation result storage means stores the evaluation result of each subcategory for each of a plurality of subcategories divided into a plurality of categories. The mapping rule storage means stores a mapping rule for each evaluation item to be extracted in each subcategory belonging to the plurality of categories, and the mapping processing procedure is mixed for each subcategory belonging to any category. The input sentence described above is extracted and stored using the corresponding category mapping rule.

  In the invention according to claim 28, in the above invention, the computer further executes a reception procedure for accepting a category designation for designating which category of the input sentence, and the mapping processing procedure includes the mapping rule Of the mapping rules for each category stored in the storage means, extraction and storage are performed using only the category mapping rules relating to the category designation received by the reception procedure.

  In the invention according to claim 29, in the above invention, the computer further executes a determination procedure for determining which category the input sentence is associated with, and the mapping processing procedure is stored in the mapping rule storage means. Of the mapping rules for each category, extraction and storage are performed using only the category mapping rules determined by the determination procedure.

  The invention according to claim 30 is an evaluation apparatus that extracts and evaluates evaluation items from an input sentence described in a natural language, and the evaluation obtained for the evaluation items of each category for each of a plurality of categories Evaluation result storage means for storing results, and for each evaluation item to be extracted in each category, an analysis result assumed as a result of natural language analysis for a sentence including the evaluation item, an evaluation result for the evaluation item, In addition, the mapping rule storage means for storing the mapping rules each defining the storage destination when the evaluation result is stored in the evaluation result storage means, and the input sentence mixedly described for each category are subjected to natural language analysis. From the obtained analysis results, the evaluation items of each category are extracted using the mapping rules stored in the mapping rule storage means, and the extracted And mapping processing means for storing an evaluation result corresponding to the evaluation item storage location corresponding in the evaluation result storage means, characterized by comprising a.

  The invention according to claim 31 is an evaluation program for causing a computer to execute an evaluation method for extracting and evaluating evaluation items from an input sentence described in a natural language. The result of the natural language analysis for the sentence containing the evaluation item for each evaluation item to be extracted in each category A mapping rule that holds a mapping rule storage unit that stores an analysis result assumed as an evaluation result, an evaluation result for the evaluation item, and a mapping rule that defines a storage destination when the evaluation result is stored in the evaluation result storage unit From the analysis result obtained by the natural language analysis of the input procedure described in the mixed procedure for each of the holding procedure, A mapping processing procedure for extracting the evaluation items of each category using the mapping rule stored in the ping rule storage unit, and storing the evaluation result corresponding to the extracted evaluation item in the corresponding storage destination in the evaluation result storage unit; Is executed by a computer.

  According to the invention of claim 1, 14 or 22, the contents of the input sentence described in a mixed manner for a plurality of subcategories are understood, the evaluation items of each subcategory are extracted, and the evaluation results for the extracted evaluation items are displayed for each subcategory. Since it is mapped to the database (evaluation result storage means), existing documents with multiple subcategories are used effectively (immediate use and reuse), and effective information in each subcategory is general-purpose from these extensive descriptions. Thus, it is possible to implement a timely, efficient and quality evaluation operation for each sub-category. Moreover, even if the input sentences have the same expression, if they match the mapping rules of each subcategory, they are extracted as evaluation items for each subcategory. Therefore, from the same sentence, a database of multiple subcategories (evaluation result storage means) is multifaceted. Mapping is also possible. Furthermore, since the function change and extension can be easily performed by correcting and adding the mapping rule, it is possible to realize an evaluation work with excellent maintainability and expandability.

  According to the invention of claim 2, 15 or 23, since the format of the mapping rule is made common among the sub-categories, the common mapping process (extraction and storage) can be executed in each sub-category, and the evaluation work Can be shared.

  According to the invention of claim 3, 16 or 24, since the condition part (analysis result) is shared on one mapping rule, the execution part (evaluation result and storage destination) of each category is specified together. In addition to improving the creation efficiency and storage efficiency of mapping rules, it is also possible to improve the evaluation efficiency of each category.

  According to the invention of claim 4, 17 or 25, since the evaluation result mapped to the database (evaluation result storage means) of each subcategory is further mapped to an arbitrary evaluation table, an arbitrary evaluation table ( It is possible to easily create a skill evaluation table in conformity with ITSS (IT Skill Standard) formulated by the Ministry of Economy, Trade and Industry.

  According to the invention of claim 5, 18 or 26, since a comprehensive evaluation table is created from the evaluation results of a plurality of subcategories, it is also possible to perform comprehensive evaluation within the category that supervises each subcategory. .

  In addition, according to the invention of claim 6, 19 or 27, since there are mapping rules for a plurality of categories, for example, input sentences for a plurality of categories such as carrier-related, consumer monitor-related, and training attendance-related are processed. Thus, it is possible to make a universal evaluation in each category. In addition, it is possible to perform the same processing for an arbitrary category, and it is also possible to simultaneously process a plurality of categories.

  According to the invention of claim 7, 20 or 28, since only the mapping rule of the category related to the category designation is used, the mapping process (extraction) is not referred to, and the mapping rule of another category is not referred to, and it is surely efficient. Mapping processing can be performed.

  According to the invention of claim 8, 21 or 29, only the determined category mapping rule is used. Therefore, the mapping rule (extraction) does not refer to the mapping rule of other categories, and the category is not designated by the user. However, an efficient mapping process can be performed.

  Further, according to the invention of claim 9, since the syntax / semantic structure assumed from the sentence including the evaluation item is defined in the mapping rule, by performing the syntax / semantic analysis on the input sentence, Evaluation items can be accurately extracted.

  Further, according to the invention of claim 10, since the mapping rule defines the phrase list (arrangement of morphemes) assumed from the sentence including the evaluation item, each subcategory can be obtained by performing morphological analysis on the input sentence. It is possible to accurately extract the evaluation items. In addition, since the evaluation items can be extracted only by morphological analysis without performing syntax / semantic analysis that takes relatively long processing time, efficient mapping processing (extraction) can be performed. Furthermore, even when syntactic / semantic analysis is performed, if the phrase list that appears in the input sentence is typical, only the phrase list can be used, even if the mapping rules are not defined with a complicated dependency structure. Mapping rules can be created easily and efficiently.

  In addition, according to the invention of claim 11, since mapping is performed using a so-called question / answer table, it is easy to achieve consistency with a conventional system that inputs the answer of the question / answer table and stores it in the database, It becomes possible to easily shift from the conventional system.

  Further, according to the invention of claim 12, since the shortage of the evaluation result obtained from the input sentence is supplemented using the question / answer table, the evaluation result for the necessary evaluation item can be easily supplemented, Evaluation can be performed without being affected by the quality of the input sentence.

  According to the invention of claim 13, since the evaluation work is performed using voice recognition, the input sentence can be easily input, and more effective information can be accurately extracted from a wider range of contents. Is possible.

  Further, according to the invention of claim 30 or 31, the content of the input sentence described in a mixed manner for a plurality of categories is understood, the evaluation items of each category are extracted, and the evaluation results for the extracted evaluation items are displayed for each category. Since it is mapped to the database (evaluation result storage means), for example, the input sentence mixed with multiple categories such as career-related, consumer monitor-related, training-related, etc. is processed, and each category is determined from these extensive descriptions. It is possible to extract effective information in the general-purpose, and to implement a timely and efficient evaluation work with good efficiency and quality in each category. In addition, it is possible to perform the same processing for an arbitrary category, and it is also possible to simultaneously process a plurality of categories. In addition, since the function can be easily changed or expanded by correcting or adding the mapping rule, it is possible to realize an evaluation operation with excellent maintainability and expandability.

  Exemplary embodiments of an evaluation apparatus, an evaluation method, and an evaluation program according to the present invention will be described below in detail with reference to the accompanying drawings. In the following, the main terms used in the present embodiment, the outline and features of the general-purpose evaluation apparatus according to the present embodiment, the configuration of the general-purpose evaluation apparatus, and the flow of the general-purpose evaluation processing will be described in order, and finally various aspects of the present embodiment will be described. A modified example will be described.

[0: Explanation of terms]
First, main terms used in this embodiment will be described. The “skill” used in the present embodiment is one of the evaluation items to be evaluated for employment, personnel, etc., such as management ability and development ability. Further, “motivation” used in the present embodiment is one of the evaluation items to be evaluated for job hunting, personnel affairs, and the like, such as innovation-oriented and specialized / full-time-oriented strength. Furthermore, “mental” used in the present embodiment is one of evaluation items to be evaluated for job hunting, personnel affairs, etc., such as mental fatigue and satisfaction with the current situation.

  The “carrier-related document (corresponding to“ input sentence ”described in claims”) ”used in the present embodiment is, for example, in the same document such as a business career record, career survey form, and personnel evaluation form. In addition, it is a document that contains information about multiple subcategories such as holding skill information, motivation / application information, and mental information.

  In other words, the above-mentioned “skills”, “motivation”, and “mental” correspond to “subcategories” described in the claims, and the scope of “career-related” that supervises these subcategories is included in the claims. Corresponds to the “category” described. In addition, “skill items”, “motivation items”, and “mental items” used in this embodiment correspond to “evaluation items of each subcategory” described in the claims.

  In the following, “syntax / semantic structure” is expressed as “dependency structure” or “syntax analysis result”, and “morpheme structure” is expressed as “morpheme analysis result”, “morpheme string”, “word string”, etc. Each of which has the same meaning.

[1: Overview and features of general-purpose evaluation equipment]
Next, the outline | summary and characteristic of the general purpose evaluation apparatus based on a present Example are demonstrated using FIGS. 1-3. FIG. 1 is a block diagram illustrating a configuration of a general-purpose evaluation apparatus according to the present embodiment, and FIGS. 2 and 3 are diagrams for explaining an overview of a general-purpose evaluation process according to the present embodiment.

  The general-purpose evaluation apparatus 200 according to the present embodiment has an outline of a career-related document in which evaluation items of a plurality of subcategories such as skill items, motivation items, and mental items are mixed, and skill evaluation and motivation evaluation in each subcategory. In addition to general evaluation, mental evaluation is performed on each sub-category, and a skill evaluation table, motivation evaluation table, mental evaluation table, and comprehensive evaluation table are output.

  In other words, for example, input of career-related documents such as “I was in charge of management of 10 people and developed an accounting system. I am dissatisfied because the specialized field is different. I want to improve the specialized ability of the network next year.” When a sentence is entered, skill items are extracted and evaluated from “I managed 10 people and developed an accounting system.” From “I am dissatisfied because the specialized fields differ.” Mental items are extracted and evaluated, and “I want to improve my network expertise next year” is to extract and evaluate motivation items.

  Therefore, in the general-purpose evaluation apparatus 200 according to the present embodiment, as illustrated in FIG. 3, the evaluation results obtained for the evaluation items of each subcategory are stored for each of a plurality of subcategories such as skill, motivation, and mental. A database (skill DB, motivation DB, mental DB) is provided.

  Further, in the general-purpose evaluation apparatus 200 according to the present embodiment, as illustrated in FIG. 2, for each subcategory of skill, motivation, and mental, the question relating to the evaluation item of each subcategory and the answer to the question were obtained. Question / answer table (skill question / answer table, motivation question / answer table, mental) for storing the evaluation results and the storage locations when the evaluation results are stored in the database (skill DB, motivation DB, mental DB) Question / answer table).

  Furthermore, in the general-purpose evaluation apparatus 200 according to the present embodiment, as illustrated in FIG. 2, for each evaluation item to be extracted in each subcategory, it is assumed as a result of syntax / semantic analysis for a sentence including the evaluation item. Stored in the question / answer table (skill question / answer table, motivation question / answer table, mental question / answer table) Are provided with mapping rules (skill mapping rules, motivation mapping rules, mental mapping rules) respectively defined.

  Then, in the general-purpose evaluation apparatus 200 according to the present embodiment, as illustrated in FIGS. 2 and 3, the evaluation items of each subcategory are extracted from the analysis result of the carrier-related document using the mapping rule, and the extracted evaluation The evaluation result corresponding to the item is stored in the corresponding storage location of the question / answer table, and the evaluation result is also stored in the database (skill DB, motivation DB, mental DB) stored in advance in the question / answer table. To map.

This will be explained with a specific example.
If <dependency structure> =
[Keywords (in charge | implementation)]
|-(Target) [Keyword (Management | Management)] + [Keyword ()]
|-(Target) [Part of speech (Numeric)] + [Keyword (People)] + [Keyword (No)]
Then <application processing> = QA_SKILL (2) & (answer = [part of speech (numerical value)])
As a result of matching the input sentence “10 people in charge of management” with the skill mapping rule of “QA_SKILL (2)” in the execution part, the document whose condition part is matched is Associated with the second question in the answer sheet.

  In other words, by applying such a mapping rule, as the answer to the second question “How many people are managed?” In the skill question / answer table, check the “Yes / No” column in the “Answer” column of the question / answer table, Further, “10” (the part of speech of the headword “10” is a numerical value) in the input sentence is stored as an answer by “(answer = [part of speech (numerical value)])” of the execution unit. Furthermore, using the association with the data item of the database described in the response processing column of the question / answer table, the data of skill category related to project management (PRJ management) of the skill database is obtained from the answer of “10 management”. The skill item is stored in the item.

  As described above, when the input sentence matches the skill mapping rule, not only the skill item is extracted but also associated with the data item (= data storage position) of the skill database through the skill question / answer table. However, this association means “which skill item (skill category) and which skill level”. That is, at the stage matching the skill mapping rule, skill items are extracted, the skill level is evaluated, and further, the position in the skill database where data should be stored (associated) is known. This means that skill evaluation has been performed for one skill at the stage where the rule is matched.

  In the general-purpose evaluation apparatus 200 according to the present embodiment, the mapping rules as described above are provided for each of the skill, motivation, and mental subcategories. For this reason, when one career-related document is input, the above-described evaluation is performed for each subcategory of skill, motivation, and mental, and a skill DB, a motivation DB, and a mental DB are created.

  Furthermore, in the general-purpose evaluation apparatus 200 according to the present embodiment, as illustrated in FIG. 3, an evaluation table for each subcategory is created from the evaluation results (skill DB, motivation DB, mental DB) for each subcategory, and overall evaluation is performed. Create a table. That is, a skill evaluation table, a motivation evaluation table, a mental evaluation table, and a comprehensive evaluation table are output.

  As described above, the general-purpose evaluation apparatus 200 according to the present embodiment understands the contents of the input sentence described in a mixed manner for a plurality of subcategories, extracts the evaluation items of each subcategory, and outputs the evaluation results for the extracted evaluation items. Since mapping to the database of each subcategory, existing documents with multiple subcategories are used effectively (immediate use, reuse), and useful information in each subcategory is extracted generically from these extensive descriptions, In each subcategory, it is possible to implement evaluation work with good timeliness and efficiency and quality in general.

  In other words, just enter one career-related document (business history, career survey form, personnel evaluation form, etc. that contains information on multiple subcategories such as holding skill information, motivation / application information, mental information, etc. in the same document) Since the related information of each subcategory is simultaneously mapped (extracted and stored), skill evaluation, motivation evaluation, and mental evaluation can be performed simultaneously.

  By the way, the survey / evaluation is usually performed within a specific category (for example, carrier-related, consumer monitor-related, etc.). Documents in each category are often described in a predetermined style (auxiliary verb, modifier type, etc.) and technical terms corresponding to the category. In addition, when one category is subdivided into subcategories, there is a tendency that certain differences in style and technical terms are seen between subcategories. In particular, when information on a plurality of subcategories is described in the same document, the tendency is remarkable. This may be because the writer has given a certain difference in style and expression so that the reader can clearly distinguish which subcategory each sentence is described.

  For example, in career-related categories, career-related terms such as skill names and job titles are generally used, and the style is also a fact or the future. Tend to be. For example, if this is separated into a sub-category of “skills” and a sub-category of “motivation”, an auxiliary verb or adverb representing intention or future is likely to be attached to the motivation information, such as “Tai” or “Next year”. There is a tendency. A specific tendency is also observed in documents other than the carrier-related category.

  In the mapping rules used in the general-purpose evaluation apparatus 200 according to the present embodiment, paying attention to this point, expression patterns that are likely to appear in each subcategory are ruled as mapping rule <condition part> (dependency structure = syntax structure). And since the method of implementing mapping rules using this expression pattern as a clue is based on the description characteristics of the document (that it is easy to understand what is said), create a rule. Not only is it efficient, it also has a certain effect on improving the quality of information extraction during processing.

[2: Configuration of general-purpose evaluation device]
Next, the configuration of the general-purpose evaluation apparatus 200 according to the present embodiment will be described with reference to FIGS. 1 and 4 to 23. As shown in FIG. 1, the general-purpose evaluation apparatus 200 according to the present embodiment includes a natural language processing unit 201, a mapping rule storage unit 202, a matching unit 203, a rule editing unit 204, an application processing unit 205, It includes a question / answer information storage unit 206, an answer complement processing unit 207, a mapping unit 208, an evaluation result storage unit 209, an evaluation analysis unit 210, an evaluation table creation unit 211, and a conversion rule storage unit 212.

  Among these, the natural language processing unit 201 is a processing unit that performs syntax analysis and semantic analysis by inputting carrier-related documents written in natural language such as business career records, career survey forms, and personnel evaluation forms. 4 to 6 are diagrams illustrating examples of natural language processing by the natural language processing unit 201. FIG.

  As shown in FIG. 4, the natural language processing unit 201 inputs, for example, a sentence “I was in charge of management of 10 people and developed an accounting system”, performs morphological analysis using an electronic dictionary, As a result of the analysis, a word string “10 / person / of / management / of / in charge / do / do // accounting system / development / ta /./” is obtained. In addition, the natural language processing unit 201 adds grammar / semantic information such as part of speech, utilization, utilization form, and meaning to each word delimited by “/”.

  For example, for the word “10”, “number” as part of speech, “unutilized” as utilization, and “unutilized” as utilization form are added as grammar / semantic information. The word “people” is interpreted as two types of morphemes (multiple parts of speech), and one morpheme (word) is “noun” as part of speech, “no use” as usage, and “no use” as usage. “Practical”, “Person” as grammatical / semantic information as semantic elements, and other morphemes (words) as “unit noun” as part of speech, “unutilized” as practical, “unutilized” as semantic, "Number / unit, person" is added as grammar / semantic information.

And by performing syntax analysis and semantic analysis using analysis rules for this morphological analysis result,
[Keyword (development)] + [Keyword (ta)]
|-(Target Case) [Keyword (Accounting System)] + [Keyword ()]
|
|-(Optional) [Keyword (do)]
|-(Target) [Keyword (in charge)] + [Keyword ())]
|-(Target) [Keyword (Management)] + [Keyword ()]
|-(Target) [Keyword (10)] + [Keyword (People)] + [Keyword ())
Get the result of parsing and semantic analysis.

  Here, “|” and “|-” are notation symbols for indicating that a node (phrase) and a node (phrase) are linked by a predetermined dependency relationship such as an upper-lower relationship. Further, “arbitrary” displayed with “()” immediately after these symbols indicates that the relationship between the phrases is arbitrary, and “target case” indicates that the phrase is the target case. These correspond to case relationships (main case, purpose case, opponent case, etc.) and attribute relationships (subject, subject, possession, etc.) handled in natural language processing.

  In addition to the information shown above, the information of each node (phrase) in the parsing result holds grammatical and semantic information such as part of speech, utilization, meaning, etc., but is displayed as above for convenience. did. Depending on the space of the explanation, in some cases, it may be displayed only with headwords, and information indicating the relationship between nodes (“target case”, “arbitrary”, etc. in () above) may be omitted. However, it is assumed that there is no difference from the case where the stored data is completely expressed.

For example, the analysis result may be expressed as follows using only a headword.
[Develop] + [Ta]
|-[Accounting system] + [O]
|
|-[Do]
|-[Responsible] + [O]
|-[Management] + [no]
|-[10] + [People] + [no]

  By the way, in the above description, a case where the sentence “I was in charge of management of 10 people and developed an accounting system” was described as an example of processing of the natural language processing unit 201. When a sentence such as “I want to improve the specialized ability of the network” is input, the same processing is performed as illustrated in FIGS. 5 and 6.

  The carrier-related document described above is directly input through an input device such as a keyboard, a mouse, or a microphone, or is recorded on a computer-readable recording medium such as a hard disk, a flexible disk (FD), a CD-ROM, an MO, or a DVD. It is recorded and input, or input via a network such as the Internet.

  In other words, various input modes can be adopted, but an input mode (so-called so-called input information is created by inputting the carrier-related document as voice information by allowing the evaluated person to speak freely and recognizing it as voice information. In the case of adopting (hearing), it is possible to easily input an input sentence, and it is possible to accurately extract more effective information from a wider range of contents.

  Returning to FIG. 1, the mapping rule storage unit 202 is a storage unit that stores a mapping rule for each evaluation item to be extracted in each subcategory, and includes a skill mapping rule 202a, a motivation mapping rule 202b, and a mental mapping rule 202c. Remember each one. FIG. 7 is a diagram illustrating a mapping rule format and a configuration example of the skill mapping rule 202a, and FIG. 8 is a diagram illustrating a configuration example of the motivation mapping rule 202b and a configuration example of the mental mapping rule 202c.

  As shown in “Format” in FIG. 7, the mapping rule has a format composed of a condition part and an execution part such as “If <dependency structure> Then <application process>”. Here, the <dependency structure> of the condition part has the same data structure as the “syntax / semantic relationship (dependency structure)” obtained by the syntax analysis / semantic analysis. The <application process> of the execution unit is a process of associating with the question / answer table.

Then, as shown in the skill mapping rule 202a in FIG. 7, for example, from the natural language sentence “10 people in charge of management” to the answer to the second question in the skill question / answer table “how many people are managed”? The leading skill mapping rule 202a is
If <dependency structure> =
[Keywords (in charge | implementation)]
|-(Target) [Keyword (Management | Management)] + [Keyword ()]
|-(Target) [Part of speech (Numeric)] + [Keyword (People)] + [Keyword (No)]
Then <application processing> = QA_SKILL (2) & (answer = [part of speech (numerical value)])
It becomes.

  Here, “|” indicates an OR condition. That is, “in charge | implement” is “in charge” or “implementation”. In this example, the relationship between the nodes is limited to “(target)”, but the scope of application can be expanded to an arbitrary relationship and described as “(optional)”.

  In the above skill mapping rule 202a, “QA_SKILL (2)” of the execution unit indicates that the document having the matched condition unit is associated with the second question in the skill question / answer table described later. Therefore, when this skill mapping rule 202a is applied, the question / answer table is used as an answer to the second question “how many people are managed” in the skill question / answer table described later by “QA_SKILL (2)” of the execution unit. In the “Respondence presence / absence” field, “Yes” is checked, and “(answer = [part of speech (numerical value)]”) of the execution part is used, and “10” in the input sentence as the answer (part of speech of “10”) Is a numerical value) (see FIG. 10).

  As the answer information, as in the above-mentioned answer, it is possible to hold any word or phrase in the input sentence or the entire input sentence (“10 people in charge of management”) as a skill item. For questions in the YES / NO format, a rule such as “(answer = Y)” or “(answer = N)” can be described to process the question as if it was answered as YES / NO. .

  In the above description, the skill mapping rule 202a has been described as an example of the mapping rule. However, the motivation mapping rule 202b and the mental mapping rule 202c are also defined in a common format as illustrated in FIG. In other words, by making the mapping rule format common among the sub-categories, it is possible to execute common mapping processing (extraction and storage) in each sub-category, and to share the evaluation work.

  Here, information that can be specified in the condition part of the mapping rule will be described. FIG. 9 is a diagram for explaining information that can be specified in the condition part of the mapping rule. As shown in “Each node” in the figure, as the information that can be specified for each node in the condition part, that is, the part surrounded by [], the meaning (superordinate concept) is specified in addition to the part of speech or headword You can also For example, [headword (personnel system) part of speech (noun) meaning (business system)] indicates that the personnel system is one of business systems. Further, for example, [headword (LAN) part of speech (noun) meaning (network)] indicates that the LAN is one of the networks.

  In addition, as shown in “Inter-node information” in the figure, information between nodes in <dependency structure> includes case relationships (major case, purpose case, opponent case, etc.) and attribute relationships (subject, subject, possession, etc.) Can also be specified.

For example,
If <dependency structure> =
[Keywords (in charge | implementation)]
|-(Target) [Keyword (Management | Management)] + [Keyword (of)]
|-(Target) [part of speech (numerical value)] + [entry word (person)] + [entry word (no)]
Then <application processing> = QA_SKILL (2) & (answer = [part of speech (numerical value)])
Then, “Management | Management” means “object” of “in charge” in the sense of “responsible for management”, etc. “● Person” means “target” of “management” in the meaning of “● manage people”, etc. Is.

  In addition, since a plurality of OR conditions including “headword” can be described in each node, it is possible to deal with fluctuations in the notation. In addition, it is possible to deal with fluctuations in the notation by designating with “meaning”, and it is possible to deal with without increasing the number of skill mapping rules. Furthermore, arbitrary words can be matched by using “*” as a world card.

  Thus, the <dependency structure> described in the condition part of the mapping rule has a data structure in the same format as the dependency structure of the syntax / semantic analysis result. Similarly, <word list> described in a condition part of a mapping rule (see FIG. 26) described later also has a data structure in the same format as the word string of the morphological analysis result. For this reason, it becomes possible to accurately extract the evaluation items of each subcategory by understanding the contents of the input sentence.

  Returning to FIG. 1, the matching unit 203 receives the result of syntax analysis / semantic analysis from the natural language processing unit 201, and stores each mapping rule (skill mapping rule 202a, motivation mapping rule 202b, It is a processing unit that performs matching processing with the mental mapping rule 202c).

  Specifically, the matching unit 203 compares the result of syntax analysis / semantic analysis performed by the natural language processing unit 201 with the condition part of each mapping rule, and the dependency structure of the result of syntax analysis / semantic analysis is a condition. Search for a mapping rule that matches the dependency structure of the part. The matching unit 203 performs matching processing between the syntax analysis / semantic analysis result of each sentence of the input document and the condition part of the mapping rule, so that the evaluation items (skill items) of each category from the document written in natural language. , Motivation items, mental items) can be extracted.

  Here, the matching unit 203 performs mapping of each subcategory (skill mapping rule 202a, motivation) stored in the mapping rule storage unit 202 based on the result of syntax analysis / semantic analysis sequentially output from the natural language processing unit 201. Matching processing is performed with all of the mapping rule 202b and the mental mapping rule 202c). However, as described above, since the format of the mapping rule is common among the sub-categories, the matching unit 203 can execute a common matching process without being conscious of each sub-category.

  Returning to FIG. 1, the rule editing unit 204 is a processing unit that edits the mapping rule storage unit 202. Specifically, each mapping rule is added to the mapping storage unit 202 and stored in the mapping storage unit 202. Modify or delete mapping rules.

  The application processing unit 205 is a processing unit that processes the execution unit of the mapping rule searched by the matching unit 203. The question / answer information storage unit 206 stores a question / answer table (skill question / answer table 206a, motivation question / answer table 206b, mental question / answer table 206c) of each subcategory. The answers are stored in association with each other. In other words, the application processing unit 205 writes the answer created by processing the mapping rule execution unit in the question / answer table of the question / answer information storage unit 206.

  The entry to the question / answer table (application of mapping rules) will be described. FIG. 10 is a diagram showing an example of rule application from the skill mapping rule 202a to the skill question / answer table 206a, and FIG. 11 is a diagram showing an example of rule application from the motivation mapping rule 202b to the motivation question / answer table 206b. FIG. 12 is a diagram showing a rule application example from the mental mapping rule 202c to the mental question / answer table 206c.

  For example, when the skill mapping rule 202a illustrated in FIG. 7 is searched (matched) by the matching process of the matching unit 203, as illustrated in FIG. 10, the application processing unit 205 includes the hit skill mapping rule 202a. In response to “QA_SKILL (2)” of the execution unit, check “Yes” in the “Respondence presence / absence” column of the question / answer table as an answer to the second question “How many people are managed” in the skill question / answer table 206a, Further, “10” (the part of speech of the headword “10” is a numerical value) in the input sentence is stored as an answer by “(answer = [part of speech (numerical value)])” of the execution unit.

  Note that “Table_SKILL (1,4) & Table_SKILL (1,5)” etc., which is described in the second column in the correspondence processing column of the skill question / answer table 202a illustrated in FIG. Indicates that the extracted skill items (response contents) are stored in association with the data items specified by (1,4) and (1,5). Similarly, “Table_SKILL (3, *)” in the third response process is (3, *) in the skill database ([*] represents an arbitrary skill level (= line position)) ) Indicates that the data item is stored in association with the data item.

  By the way, although rule application of the skill mapping rule 202a has been described here as an example of rule application by the application processing unit 205, rule application by the motivation mapping rule 202b and the mental mapping rule 202c is also exemplified in FIGS. In addition, the same processing is executed. That is, as described above, since the format of the mapping rule is common among the sub-categories, the application processing unit 205 can execute the common rule application process without being conscious of each sub-category. .

  The question / answer table can hold a plurality of answers to one question. For example, when the reply contents (skill evaluation result) “development of accounting system” and “development of personnel system” are obtained from the input sentence, the skill question / answer table of FIG. It can be held as an answer to the third question 206a "what system has been developed".

  Returning to FIG. 1, among the questions in the question / answer table stored in the question / answer information storage unit 206, the answer complement processing unit 207 complements the answer to the question for which no answer was created by the application processing unit 206. It is a processing unit. Specifically, a question / answer stored in the question / answer information storage unit 206 is obtained from the evaluated person as needed by, for example, outputting the question itself to the evaluated person by voice. Fill in the table. In this way, by supplementing the shortage of evaluation results obtained from the input sentence using the question / answer table, it is possible to easily supplement the evaluation results for the required evaluation items, which affects the quality of the input sentence. It becomes possible to evaluate without receiving.

  In addition, the answer supplement processing unit 207 can also enter answers in the question / answer table only for essential questions among the questions for which no answer was created by the application processing unit 206. That is, as illustrated in FIG. 10 to FIG. 12, information that distinguishes “questions for which it is essential to obtain answers” and “questions for which it is optional to obtain answers” for questions in the question / answer table (required / By storing (optional) in the question / answer table, even if there are insufficient evaluation items (answers, evaluation results) in the input sentence, only the required questions can be supplemented. . As a result, even if there are a large number of questions, it is possible to efficiently evaluate only with essential information.

  Returning to FIG. 1, the mapping unit 208 creates each database (skill DB 209a, motivation DB 209b, mental DB 209c) of the evaluation result storage unit 209 from the information of the question / answer table stored in the question / answer information storage unit 206. It is a processing unit. The evaluation result storage unit 209 is a database that stores evaluation results in each subcategory of the person to be evaluated. That is, the mapping unit 208 associates the answer contents of the question / answer table with the data items of each database according to the method described in the corresponding processing column of the question / answer table.

  The association process (mapping process) will be specifically described. FIG. 13 is a diagram showing an example of mapping from the skill question / answer table 206a to the skill DB 209a, FIG. 14 is a diagram showing an example of mapping from the motivation question / answer table 206b to the motivation DB 209b, and FIG. It is a figure which shows the example of a mapping from the mental question and answer table 206c to mental DB209c.

  For example, as illustrated in FIG. 13, in the correspondence processing column of the skill question / answer table 206a, the extracted skill item (response content) is a data item corresponding to which skill level of which skill category in the skill database 209a. The information indicating whether or not to be stored is described, and by storing the skill item (answer content) in the skill database 209a, the skill level evaluation result in one skill category is stored.

  In addition, in the corresponding processing column of the question / answer table, it is possible to dynamically control the data items to be associated according to the answer contents (that is, it is possible to describe conditional branching in the corresponding processing column). For example, as illustrated in FIG. 13, if the response content is “10 (person management)”, it is associated with a skill level 3 data item, and if it is “50 (person management)”, it is associated with a skill level 4 data item. Can be done.

  On the other hand, for example, in the skill database 209a, as illustrated in FIG. 13, the horizontal direction (column) means the type of skill item (skill category that is classified broadly by job type and further classified by job type), Data in a two-dimensional matrix format in which the vertical direction (row) means the degree of each skill level is held for each person to be evaluated.

  Further, the first number in parentheses in the above-mentioned question / answer table corresponding processing column is, for example, the position in the horizontal (column) direction in the skill database 209a, which is a skill category provided for each occupation type. Point to. The second number in parentheses is the position in the vertical (row) direction in the skill database 209a, which indicates the skill level in the skill category.

  In addition, when the above-described skill mapping rule 202a is matched, not only the skill item is extracted, but also the data item (= data storage position) of the skill database 209a via the skill question / answer table 206a. Although this association is made, this association means “which skill item (skill category) is at which skill level”. That is, at the stage matching the skill mapping rule 202a, skill items are extracted, the skill level is evaluated, and further, the position in the skill database 209a where data should be stored (associated) is known. . This means that skill evaluation has been performed for one skill when the rule matches.

  Then, as a mapping process for the database as described above, the mapping unit 208 uses an association with the data item of the database described in the corresponding process column of the question / answer table, and answers (skill items, etc.) ) In the corresponding position in the database. For example, as illustrated in FIG. 13, a skill item is stored in the skill category data item related to project management (PRJ management) in the skill database 209 a from the answer “management of 10 people”. From the reply “Developed”, the skill item is stored in the data item of the skill category related to application development (AP development) in the skill database 209a.

  By such mapping processing, for example, skill items related to the skill category (AP development) related to the application development experience of the evaluator and the skill category related to project management (PRJ management) are stored in the data items that match those skill levels. It is possible.

  Different evaluation items may be stored at the same position in the database. For example, when an input sentence that an accounting system and a personnel system are developed is evaluated, it is assumed that “development of accounting system” and “development of personnel system” are stored in the same data item of the skill DB 209a. Thereby, when a plurality of skill items are stored in the same data item, the skill level can be interpreted higher.

  In the above description, mapping to the skill DB 209a has been described as an example of mapping by the mapping unit 208, but mapping to the motivation DB 209b and the mental DB 209c is also executed by the same processing as illustrated in FIGS. That is, since the format of the correspondence processing column of the question / answer table is common among the sub-categories (see FIGS. 13 to 15), the mapping unit 208 performs common mapping processing without being conscious of each sub-category. It is possible to execute.

  Returning to FIG. 1, the evaluation result analysis unit 210 outputs a result of analyzing and evaluating the skill, motivation, and mentality of the evaluated person based on the evaluation result of each subcategory stored in the evaluation result storage unit 209. Part. Further, the evaluation result analysis unit 210 performs aggregation and analysis of skills evaluation results, motivation evaluation results, and mental evaluation results for the entire evaluated person stored in the evaluation result storage unit 209, and displays the analysis results. .

  The evaluation table creation unit 211 is a processing unit that converts the evaluation result stored in the evaluation result storage unit 209 into an evaluation table having an arbitrary format and outputs the evaluation table. The conversion rule storage unit 212 is a storage unit that stores a conversion rule for converting the evaluation result stored in the evaluation result storage unit 209 into an evaluation table of an arbitrary format, and includes a skill conversion rule 212a and a motivation conversion rule. 212b, mental conversion rule 212c, and comprehensive evaluation rule 212d are stored. That is, the evaluation table creation unit 211 outputs a skill evaluation table, a motivation evaluation table, a mental evaluation table, and a comprehensive evaluation table in an arbitrary format based on the rules stored in the conversion rule storage unit 212.

  Such conversion rules (skill conversion rule 212a, motivation conversion rule 212b, mental conversion rule 212c) will be described. FIG. 16 is a diagram illustrating a configuration example of the conversion rule format and the skill conversion rule 212a, and FIG. 17 is a diagram illustrating a configuration example of the motivation conversion rule 212b and a configuration example of the mental conversion rule 212c.

  As shown in “Format” in FIG. 16, in the condition part of each conversion rule, the row and column of each database (skill DB 209a, motivation DB 209b, mental DB 209c) in the evaluation result storage unit 209 (positions of data items in the database) If necessary, application conditions for evaluation items can be specified, and the execution unit matches the evaluation item (evaluation result) at the position described in the condition part with a predetermined position in the evaluation table of any format. Then, a process is described in which the evaluation items held in the database are reflected as they are or stored after performing some arithmetic processing (such as point conversion for the evaluation result).

  However, it is only necessary to describe the application conditions for evaluation items. When converting, if there is a conversion rule that matches the condition part for all the rows and columns of the database, the execution part is sequentially applied. Specifically, conversion rules as exemplified in FIGS. 16 and 17 are described. As an example, the skill conversion rule 212a illustrated in FIG. 16 shows a conversion rule when the skill DB 209a is converted into an evaluation table conforming to ITSS (IT Skill Standard) established by the Ministry of Economy, Trade and Industry. It is a thing.

  Here, creation of an evaluation table based on the conversion rule as described above will be described. 18 is a diagram showing an example of conversion from the skill DB 209a to the skill evaluation table, FIG. 19 is a diagram showing an example of conversion from the motivation DB 209b to the motivation evaluation table, and FIG. 20 is a motivation evaluation table from the motivation DB 209c. FIG. 21 is a diagram illustrating an example of conversion from the mental DB 209d to the mental evaluation table.

  For example, as illustrated in FIG. 18, the evaluation table creation unit 211 applies the skill conversion rule 212a to the skill item related to PRJ management and AP development in the skill database 209a to the skill item related to ITSS project management and application specialist. By associating, a predetermined skill evaluation report based on ITSS is created.

  Also, for example, as illustrated in FIG. 19 and FIG. 20, the evaluation table creation unit 211 can apply any motivation conversion rule 212b to the motivation conversion rules 212b to convert the motivation items of the specialized / dedicated type or the innovative type into any arbitrary motivation definition system. Is associated with the motivation item in a predetermined direction. In this way, by associating the motivation DB 209b with an arbitrary motivation definition system, a predetermined motivation evaluation report or the like based on an arbitrary motivation definition system can be created.

  The creation of the evaluation table is the same when converting from the mental DB 209d to the mental evaluation table by using the mental conversion rule 212c as illustrated in FIG. In this way, by further mapping the evaluation result mapped to the database of each subcategory to an arbitrary evaluation table, it is possible to easily create an arbitrary evaluation table in each subcategory.

  By the way, the evaluation table as described above is merely an example, and may be mapped to another type of evaluation table. That is, it is not limited to a visual format such as “radar format or matrix format” illustrated in FIGS. 19 and 21, and a directional format such as “innovation type or specialized type” is also illustrated in FIGS. 19 and 21. For example, directions such as “realistic, research, artistic, social, corporate, and customary” can be adopted as appropriate.

  In the above, the case where individual skill items, motivation items, and mental items are mapped to the evaluation table has been described. However, the skill items, motivation items, and mental items of each person belonging to a certain organization are mapped to the evaluation table. May be.

  That is, in this case, not only the conversion rules for individuals but also the conversion rules for organizations for reflecting the evaluation items of a plurality of people in the evaluation table are stored in the conversion rule storage unit 212 (FIG. 17). (See the second motivation conversion rule 212b illustrated in FIG. 19), for example, an item having the highest motivation level among the motivation items of each person belonging to a certain organization (for example, reflected in the evaluation table shown in FIG. 19) Figure shows the items extracted from each person according to the conversion rule such that one of the plurality of plots is the one farthest from the center) is regarded as each person's overall motivation item and level and extracted one 19 and the evaluation table as shown in FIG.

  Note that the radar chart shown in FIG. 19 is suitable for a case where an individual's tendency is to be easily grasped visually, and the matrix chart shown in FIG. 20 is a case where the tendency of an organization is to be grasped at a time. (For example, when grasping how the motivation tendency of a plurality of people belonging to a certain department is dispersed).

  By the way, in the above, the case where the evaluation table creation unit 211 creates an evaluation table using each conversion rule has been described. However, the present invention is not necessarily limited to this, and the evaluation table creation unit 211 applies the conversion rule. Instead, the evaluation result of each database may be output as it is, that is, the skill evaluation table may be output in a format having the same skill category and skill level.

  Next, comprehensive evaluation processing by the evaluation table creation unit 211 will be described. FIG. 22 is a diagram illustrating a format and a configuration example of the comprehensive evaluation rule 212d, and FIG. 23 is a diagram illustrating an output example of an evaluation result. The comprehensive evaluation rule is a rule for collecting evaluation results for each subcategory to derive a comprehensive judgment (evaluation table). As such a comprehensive evaluation rule 212d, for example, as illustrated in FIG. 22, a predetermined message is output by combining evaluation results (for example, five-level evaluation values) output in each subcategory. Can be considered.

  When such a comprehensive evaluation rule 212d is prepared, the evaluation table creation unit 211 applies the comprehensive evaluation rule 212d after obtaining the five-level evaluation value and the evaluation message from the evaluation result of each subcategory. The comprehensive evaluation message is acquired, and a comprehensive evaluation table as illustrated in FIG. 23 is generated and output. In this way, by creating a comprehensive evaluation table from the evaluation results of a plurality of subcategories, it is also possible to perform comprehensive evaluation within the category that supervises each subcategory.

  Each evaluation table described above is output through an output device such as a monitor (or display or touch panel) or a speaker, or can be read by a computer such as a hard disk, flexible disk (FD), CD-ROM, MO, or DVD. It is recorded on a recording medium and outputted, or outputted to another external device via a network such as the Internet.

  The general-purpose evaluation apparatus 200 described above is realized by mounting the functions of the above-described units on an information processing apparatus such as a known personal computer, workstation, PHS terminal, portable terminal, mobile communication terminal, or PDA. You can also.

[3: Processing by general-purpose evaluation device]
Next, a processing procedure of the general-purpose evaluation apparatus 200 according to the present embodiment will be described. FIG. 24 is a flowchart showing a general-purpose evaluation processing procedure according to this embodiment. As shown in the figure, in the general-purpose evaluation apparatus 200, the natural language processing unit 201 inputs a carrier-related document written in a natural language and performs morphological analysis, syntax analysis, and semantic analysis (step S101).

  Then, the matching unit 203 sequentially selects the morpheme / syntax analysis results for the input sentence analyzed by the natural language processing unit 201 (step S102), and whether there is a mapping rule whose condition part matches the selected morpheme / syntax analysis result. Whether or not is determined by each sub-category (skill, motivation, mental) (step S103).

  As a result, when there is a mapping rule whose condition part matches the selected morpheme / syntax analysis result, the application processing unit 205 evaluates the evaluation item (or input sentence) from the morpheme / syntax analysis result (or input sentence). Skill items, motivation items, and mental items) are extracted and evaluated, and an answer to the question associated by the matching mapping rule execution unit is created (step S104). On the other hand, if there is no mapping rule whose condition part matches the selected morpheme / syntax analysis result, nothing is performed on the morpheme / syntax analysis result (input sentence).

  Then, it is checked whether or not all the sentences have been processed (step S105). If all the sentences have not been processed, the process returns to step S102 to process the next sentence. On the other hand, when all the sentences have been processed, it is checked whether there are unanswered questions in the question / answer table (step S106). As a result, when there is an unanswered question, the answer complement processing unit 207 performs information complement processing such as obtaining an answer from the evaluated person (step S107).

  In addition, by providing each question in the question / answer table with information that can distinguish whether it is a mandatory question or a non-essential question, even if there is an unanswered question, only the supplement to the required question is possible. It is also possible not to do so. Thereby, even if there are a large number of questions, it is possible to efficiently evaluate only with essential information.

  Then, the mapping unit 208 performs the process described in the corresponding processing column on the question / answer table stored in the question / answer information storage unit 206, whereby predetermined data in each database of the evaluation result storage unit 209 is obtained. Evaluation items (skill items, motivation items, mental items) are stored in the items (step S108). When there are a plurality of persons to be evaluated in the above-described processing, a question / answer table and data (table) in the evaluation result storage unit 209 are created for each person to be evaluated.

  As described above, when the matching unit 203 performs the matching process between the sentence analyzed by the natural language processing unit 201 and the mapping rule of each subcategory, and there is a matching mapping rule, the application processing unit is used by using the execution unit. 205 creates an answer to the question / answer table, and mapping unit 208 maps and stores the answer of the question / answer table to the data items of each database in evaluation result storage unit 209, so that the carrier written in natural language Evaluation items (skill items, motivation items, mental items) can be extracted from related documents, evaluated, and stored in the evaluation result storage unit 209.

  Then, the evaluation table creation unit 211 uses the conversion rules (skill conversion rules 212a, motivation conversion rules 212b, mental conversion rules 212c, and comprehensive evaluation rules 212d) stored in the conversion rule storage unit 212 to evaluate the results. From the evaluation results stored in 209, a skill evaluation table, a motivation evaluation table, a mental evaluation table, and a comprehensive evaluation table in an arbitrary format are created and output (steps S109 and S110).

[4: Effect of general-purpose evaluation device]
As described above, understand the contents of the input career-related documents, extract the evaluation items (skill items, motivation items, mental items) of each subcategory, and store the evaluation results for the extracted evaluation items in the database of each subcategory Effective mapping (immediate use, reuse) of existing career-related documents, so that effective information in each subcategory (skills, motivation, mental) is extracted generically from these extensive descriptions, It is possible to implement evaluation work in a subcategory in a timely manner with high efficiency and quality.

  In addition, even if the input sentences have the same expression, if they match the mapping rules of each subcategory, they are extracted as evaluation items for each subcategory, so it is possible to map from the same sentence to multiple subcategory databases. Become.

  In other words, for example, an input sentence that “work motivation has been decreasing recently” is extracted as mental information “prone to depression” and mapped to a mental database and “current job” Is extracted as motivation information that is less motivated (less motivated) and mapped to the motivation database, and is further extracted as skill information that is "low production capacity" and mapped to the skills database Multi-faceted mapping from the same sentence to multiple subcategories becomes possible.

  Furthermore, since the function change and extension can be easily performed by correcting and adding the mapping rule, it is possible to realize an evaluation work with excellent maintainability and expandability. In other words, it is possible to easily perform various evaluations simply by replacing the mapping rules. Also, as described later, the mapping rules are replaced with mapping rules of other categories (for example, carrier-related mapping rules are replaced with phenomenon analysis, user It is possible to easily evaluate each sub-category in other categories simply by replacing the mapping rules related to consumer monitoring to execute product surveys and needs analysis simultaneously.

  In addition, since the syntax / semantic structure assumed from the sentence including the evaluation item is defined in the mapping rule, it is possible to accurately extract the evaluation item of each subcategory by performing syntax / semantic analysis on the input sentence. It becomes possible.

  In addition, since mapping is performed using a so-called question / answer table, it is easy to achieve consistency with the conventional system in which the answers in the question / answer table are input and stored in the database, and the existing system is easily migrated. It becomes possible.

  Note that the evaluation table obtained by the general-purpose evaluation apparatus 200 according to the present embodiment is used for personnel evaluation, personnel allocation, personnel education, career consultant, mental counseling, etc., as illustrated in FIG. That is, FIG. 25 is a system configuration diagram showing the configuration of the system including the general-purpose evaluation apparatus 200 according to the present embodiment, but as shown in FIG. 25, business history, personnel evaluation information, application / motivation related information, Skill evaluation, motivation evaluation, and mental evaluation are performed from career related documents including mental health related information, and these evaluation results are used for personnel evaluation, personnel assignment, personnel education, and the like.

[5: Other embodiments]
Although the embodiments of the present invention have been described so far, the present invention may be implemented in various different forms other than the embodiments described above. Therefore, as shown below, different embodiments will be described by dividing into (1) to (11).

(1) Phrase-arrangement-type mapping rule In the above embodiment, the mapping rule that defines “syntax / semantic structure” assumed as a result of syntax / semantic analysis for a sentence including an evaluation item in the condition part has been described. However, the present invention is not limited to this. For example, it is assumed as a result of at least natural language analysis, such as defining a “phrase list” assumed as a result of morphological analysis for a sentence including an evaluation item in a conditional part. Any mapping rule may be used as long as the analysis result is defined in the condition part.

  FIG. 26 is a diagram illustrating a format of a word-arrangement-type mapping rule, a configuration example of a skill mapping rule, and a configuration example of a motivation mapping rule. As shown in the “format” in the figure, this mapping rule has a format of “If <word list> Then <application processing>”. Here, <word arrangement> has the same data structure as the “word arrangement (appearance pattern)” obtained by morphological analysis.

Also, as shown in the skill mapping rule 202a in the figure, for example, the skill mapping rule that leads to the answer to the question “how many people are managed” from the natural language sentence “10 people in charge of management”
If <Word List> =
[Part of speech (numerical value)] + [entry word (people)] + [entry word (no)] + [entry word (management | management)] + [entry word (no)] + [entry word (in charge | implementation)]
Then <application processing> = QA_SKILL (2) & (answer = [part of speech (numerical value)])
It becomes.

  Then, when this skill mapping rule 202a is applied, “QA_SKILL (2)” of the execution unit causes the question / answer table as an answer to the second question “how many people are managed” of the skill question / answer table 206a. “Yes” is checked in the “Respondence presence / absence” column, and the part of speech of the entry word “10” (“10”) in the input sentence is a numerical value by “(answer = [part of speech (numerical value)])”. ) Is stored.

Further, as shown in the motivation mapping rule 202b of FIG. 5, for example, a motivation mapping rule 202b that leads to an answer to a question “I want to pursue specialized ability” from a natural language sentence “I want to improve the specialized ability of the network”. Is
If <Word List> =
[Keyword (Professional skill | Technology | Skill)] + [Keyword ()] + [Keyword (Improve | Pursue | Wear | Learn | Learn)] + [Headword
Then <application processing> = QA_MOTIV (2) & (answer = Y) & (answer = "request: skill improvement")
It becomes.

  When this motivation mapping rule 202b is applied, the question / answer table is used as an answer to the second question “I want to pursue professional skills” in the motivation question / answer table 206b by “QA_MOTIV (2)” of the execution unit. "Yes / No answer" is checked as "Yes", and "Y, Desire" is stored as an answer by "(answer = Wish: skill improvement").

  In the above description, the skill mapping rule 202a and the motivation mapping rule 202b have been described with specific examples. However, the mental mapping rule 202c also adopts a mapping rule that defines a phrase arrangement as a condition part in the same manner. Can do.

  As described above, according to the mapping rule of the phrase arrangement format, the phrase arrangement (arrangement of morphemes) assumed from the sentence including the evaluation item is defined in the mapping rule. Therefore, by performing the morphological analysis on the input sentence Therefore, it is possible to accurately extract the evaluation items of each subcategory.

  Further, since the evaluation items can be extracted only by morphological analysis without performing syntax / semantic analysis that takes relatively long processing time, efficient mapping processing (extraction) can be performed. Furthermore, even when syntactic / semantic analysis is performed, if the phrase list that appears in the input sentence is typical, only the phrase list can be used, even if the mapping rules are not defined with a complicated dependency structure. Mapping rules can be created easily and efficiently.

(2) Direct format mapping In the above embodiment, the association to the question / answer table is specified as the <application process> of the execution part of the mapping rule. However, the present invention is not limited to this. Since <application processing> of the mapping rule can change the data of the association destination, instead of designating the association to the question / answer table, direct association to the evaluation result storage unit 209 (database of each subcategory) Can also be specified.

  27 and 28 are diagrams showing examples of direct format mapping. For example, as illustrated in FIG. 27, the direct form skill mapping rule 202a is defined so as to directly associate the input sentence with the skill database 209a.

In particular,
If <dependency structure> =
[Keywords (in charge | implementation)]
|-(Target) [Keyword (Management | Management)] + [Keyword ()]
|-(Target) [Part of speech (Numeric)] + [Keyword (People)] + [Keyword (No)]
Then <application processing> Table_SKILL (1,4) & Table_SKILL (1,5)
In the skill mapping rule 202a, the <application processing> is “Table_SKILL (1,4) & Table_SKILL (1,5)”, which is identified by (1,4) and (1,5) in the skill database 209a. The association to the data item to be specified is specified.

Also,
If <dependency structure> =
[Keywords (develop | implement | implement)]
|-(Target) [Keyword (Accounting System)] + [Keyword ())]
Then <application processing> Table_SKILL (3, *)
In the skill mapping rule 202a, <application processing> is “Table_SKILL (3, *)”, which is (3, *) ([*] is an arbitrary skill level (= line position) in the skill database 209a) To the data item specified in ().

  As described above, the skill database 209a can be efficiently created by directly specifying the association with the skill database 209a instead of the skill question / answer table 206a in the execution unit of the skill mapping rule 202a.

  In the above-described embodiment, the method of converting from the skill database 209a to the ITSS-compliant skill evaluation table using the conversion rule has been described. However, the skill mapping rule 202a starts the ITSS-compliant skill evaluation table. It is also possible to directly relate to a table (evaluation table) or the like of any format.

  Also, for example, as illustrated in FIG. 28, the direct form motivation mapping rule 202b is defined so as to directly associate the input sentence with the motivation database 209b.

In particular,
If <dependency structure> =
[Keyword (Improve | Pursue | Pursue | Educate | Learn)] + [Keyword (I want to)]
|-(Target) [Keyword (Professional skill | Technology | Skill)] + [Keyword ()]
|-(Arbitrary) [meaning (technical)] + [entry (no)]
Then <application processing> = Table_MOTIV (2,3)
In the motivation mapping rule 202b, the <application process> is “Table_MOTIV (2,3)”, which specifies the association with the data item identified by (2,3) in the motivation database 209b.

Also,
If <dependency structure> =
[Headword (Aspire)]
|-(Target Case) [Keyword (Research | R & D)] + [Keyword ()]
|-(Optional) [Keyword (Technical)] + [Keyword (of)]
|-(Optional) [Keyword (new | advanced | advanced)]
Then <application processing> = Table_MOTIV (8,6)
In the motivation mapping rule 202b, the <application process> is “Table_MOTIV (8,6)”, which specifies the association with the data item identified by (8,6) in the motivation database 209b.

  In this way, the motivation database 209b can be efficiently created by directly specifying the association with the motivation database 209b instead of the question / answer table by the execution unit of the motivation mapping rule 202b.

  In the above embodiment, a method for converting from the motivation database 209b to a predetermined motivation evaluation table (an evaluation table based on an arbitrary motivation definition system illustrated in FIGS. 19 and 20) using a conversion rule has been described. The motivation mapping rule 202b can be directly associated with a table (evaluation table) of an arbitrary format including the above motivation evaluation table.

  Although the direct mapping to the skill database 209a and the direct mapping to the motivation database 209b have been described here, the direct mapping to the mental database 209c is similarly processed.

(3) Phrase Extraction from Input Sentence In the above embodiment, the case where the evaluation result for the evaluation item is acquired as an answer has been described. However, the present invention is not limited to this, and a specific phrase ( It is also possible to extract a word related to the evaluation result and acquire the evaluation result as an answer together with the word.

That is, when a specific word is extracted from the syntax analysis result, the word is specified by “a variable with $” such as “$ 1, $ 2” as illustrated below.
$ 2 = [Keyword (develop | implement | implement)]
$ 1 = [Keyword (Accounting System | HR System | Business System | Application System)]
If <dependency structure> =
$ 2
|-(Target) $ 1 + [Keyword ()]]
Then <application processing> = QA_SKILL (3) & (answer = Y) & (answer = $ 1)

  In this example, if the input sentence is “I have developed an accounting system”, $ 1 corresponds to “Accounting system”, and the character string “answer = accounting system” is extracted. If the input sentence is “Developed HR system”, $ 1 corresponds to “HR system”, and the character string “answer = HR system” is extracted. In other words, the mapping rule of the evaluation engine can be extracted by associating a $ variable with an arbitrary phrase in the input sentence. In this way, it is possible to perform a higher quality evaluation by extracting a specific phrase (a phrase related to the evaluation result) from the input sentence and obtaining the evaluation result as an answer together with the phrase.

(4) Sharing of condition part in mapping rule In the above embodiment, the case where the mapping rule is defined for each sub-category of skill, motivation, and mental has been described. However, the present invention is not limited to this, and mapping is performed. A condition part in a rule can be shared by each subcategory.

  That is, it is possible to assign the same input sentence to a plurality of subcategories by simultaneously evaluating the same input sentence from a plurality of subcategories such as skill, motivation, and mental, in other words, by applying a mapping rule to the same input sentence. For example, from the input sentence “Recently, work motivation”, mentally evaluate it as a “depression” symptom, and motivation it evaluates “motivation for current work is reduced” Can do. In such a case, the condition part in the mapping rule is shared by each subcategory.

  29 and 30 are diagrams for explaining sharing of the condition part in the mapping rule. As described in the above embodiment, the mapping rule has a format in which the sentence structure is specified in the condition part and the mapping destination is specified in the execution part. However, when the condition part is shared, the execution part As a mapping destination specified in the above, a predetermined position in the skill question / answer table, a predetermined position in the motivation question / answer table, and a predetermined position in the mental question / answer table are written together.

Specifically, as shown in “Shared Mapping Rule” in FIG. 29,
If <dependency structure> =
[Headwords (decreasing | falling | falling | falling)]
|-(Main) [Headword (Motivation | Aggression | Motivation)] + [Headword (Ga)]
|-(Optional) [Keywords (Recent | Nearly | Nearly)]
Then <Applied processing> = QA_MENTAL (2) & (answer = N) & (answer = "symptom: possibility of depression"); QA_MOTIV (6) & (answer = Y) & (answer = "motivation: reduced") )
In the shared mapping rule, <application process> is “QA_MENTAL (2)” and “QA_MOTIV (6)”, and by applying this rule, it is associated with the second question of the mental question / answer table 206c and motivated. It is associated with the sixth question in the question / answer table 206b.

  Thus, by sharing the condition part (analysis result) on one mapping rule and specifying the execution part (evaluation result and storage destination) of each category together, the creation efficiency and storage efficiency of the mapping rule can be improved. It is possible to improve the evaluation efficiency of each category as well.

  Here, as the <application process> of the execution unit of the common mapping rule, the association to the question / answer table is specified, but the present invention is not limited to this, and as illustrated in FIG. Instead of specifying the association to the question / answer table, it is also possible to specify the direct association to the evaluation result storage unit 209 (database of each subcategory).

(5) Other category In the above embodiment, the case where the general-purpose evaluation process is performed for the category related to career that supervises each sub-category such as skill, motivation, and mental has been described, but the present invention is not limited to this, The same applies to other categories.

  That is, for example, as shown in FIG. 31, a consumer monitor-related document such as “the body gets cold in the toilet. Currently, an electric stove is attached. I want a heater that heats more compactly and rapidly.” From the above, it is possible to extract phenomenon information, product information that the user has, and request information, respectively, and perform phenomenon analysis, user product survey, and needs / oriented analysis, respectively.

  The “phenomenon, product, and need” referred to here corresponds to “skill, motivation, and mental” referred to in the above embodiment. That is, by configuring the general-purpose evaluation apparatus 200 as illustrated in FIG. 32, the general-purpose evaluation processing can be performed for the category “consumer monitor related” that controls each subcategory “phenomenon, product, and need”.

  In addition to such “consumer monitor related (category)”, “consumer monitor related (category)” that performs general-purpose evaluation processing from documents such as “I like sports. I want more sports programs”, The general evaluation process is the same in all categories, such as “Training related (category)” that performs general evaluation processing from a document such as “My skills are deficient. The contents were difficult to understand. I would like you to explain more clearly.” It can be carried out.

(6) Multiple categories In the above, the case where the general-purpose evaluation processing is performed in the category “career-related”, and further in the categories “consumer monitor-related (category)” and “training-related” has been described. The general-purpose evaluation apparatus 200 can be configured to process a plurality of categories instead of processing only the categories.

  That is, in this case, as illustrated in FIG. 33, in the mapping rule storage unit 202, the question / answer information storage unit 206, the evaluation result storage unit 209, and the conversion rule storage unit 212, the mapping rule and question for each category・ Has an answer sheet, database, and conversion rules. The matching unit 203 performs general-purpose evaluation processing by matching the input sentence with mapping rules of all categories (more specifically, each subcategory belonging to all categories).

  However, the mapping rule defines a condition part corresponding to the characteristics of each category, and the mapping rule searched (extracted) by the matching part 203 is matched with the category of the input sentence even if it matches with the mapping rule of all categories. Since it is limited to the mapping rule of the subcategory to which it belongs, as a result, only the evaluation table corresponding to the category of the input sentence is output. That is, it is possible to automatically obtain only the evaluation table corresponding to the category of the input sentence without particularly specifying the category.

  As described above, by having mapping rules for a plurality of categories, for example, input sentences for a plurality of categories such as career-related items, consumer monitor-related items, and training-related items are processed, and each category is evaluated universally. It becomes possible. In addition, it is possible to perform the same processing for an arbitrary category, and it is also possible to simultaneously process a plurality of categories.

  As described above, even when the general-purpose evaluation apparatus 200 is configured to process a plurality of categories, the format of the mapping rule is made common among the categories (of course, between the subcategories belonging to each category). As a result, a common mapping process (extraction and storage) can be executed in each category, and the evaluation work can be shared.

(7) Category designation / judgment Further, as described above, when processing a plurality of categories, as shown in FIG. 34, a category designation for designating which category is an input sentence is accepted. You may do it. In this case, the category selection unit 220 selects the category of the input sentence based on the category designation input via a keyboard, a mouse, or the like. The matching 203 performs the matching process using only the mapping rule for the selected category (the mapping rule for each subcategory belonging to the category).

  In this way, by using only the category mapping rules related to category designation, it is possible to reliably perform efficient mapping processing without referring to mapping rules of other categories during matching processing (extraction). .

  Instead of accepting the category designation as described above, as shown in FIG. 35, the general-purpose evaluation apparatus 200 may determine which category is the input sentence. In this case, the category determination unit 221 determines the category of the input sentence by performing so-called category determination on the input sentence. Then, the matching 203 performs the matching process using only the determined category mapping rule.

  As one method of category determination (field determination), terms having a high appearance frequency for each category are stored in the DB in advance, and the DB in which the term included in the input sentence is the most is determined as the corresponding field. Although a method is mentioned, you may determine a field | area by what kind of method other than this.

  In this way, by using only the determined category mapping rule, the mapping process of other categories is not referred to in the matching process (extraction), and the efficient mapping process can be performed without the category designation by the user. It becomes possible.

  By the way, in the above, it has been explained that only the evaluation table corresponding to the category of the input sentence can be automatically obtained without specifying the category. However, depending on the contents of the input sentence, etc. A mapping rule may be applied. However, even when such an incorrect category mapping rule is searched, it is considered that the number of erroneous mapping rule searches (number of hits) is small.

  Therefore, as illustrated in FIG. 36, after the category determination unit 222 is provided downstream and the input sentence is matched with the mapping rules of all categories, the category determination unit 222 uses the category mapping rule with a small number of hits. The mapping process may be performed based on only the mapping rules of the remaining categories.

(8) Input sentence mixed for a plurality of categories In the above-described embodiment, a case has been described in which an input sentence described mixedly for a plurality of subcategories is processed, but the present invention is not limited to this. Input sentences described in a mixed manner for a plurality of categories can be processed in the same manner.

  In this case, the evaluation result storage unit stores the evaluation results obtained for the evaluation items of each category for each of a plurality of categories. In addition, the mapping rule storage unit, for each evaluation item to be extracted in each category, an analysis result assumed as a result of natural language analysis for a sentence including the evaluation item, an evaluation result for the evaluation item, and The mapping rule which respectively specified the storage destination at the time of storing the said evaluation result in an evaluation result storage part is memorized.

  Then, the mapping unit extracts the evaluation items of each category from the analysis result obtained by performing natural language analysis on the input sentence mixedly described for each category, using the mapping rule stored in the mapping rule storage unit. The evaluation result corresponding to the extracted evaluation item is stored in the corresponding storage location in the evaluation result storage unit. That is, the same processing is realized by replacing “subcategory” described in the above embodiment with “category”.

  In this way, the contents of the input sentence described in a mixed manner for a plurality of categories are understood and the evaluation items of each category are extracted, and the evaluation results for the extracted evaluation items are stored in the database (evaluation result storage means) of each category. By mapping, for example, career-related, consumer-monitoring-related, training-related, etc. input sentences mixed in multiple categories can be processed, and effective information in each category can be generically used from these extensive descriptions. It is possible to extract and perform evaluation work with good efficiency and quality in a timely manner in each category. In addition, it is possible to perform the same processing for an arbitrary category, and it is also possible to simultaneously process a plurality of categories. In addition, since the function can be easily changed or expanded by correcting or adding the mapping rule, it is possible to realize an evaluation operation with excellent maintainability and expandability.

  In addition, input sentences described for a certain category can be processed in the same way, instead of input sentences described for a plurality of categories. In this case, timely efficiency and quality only for a certain category. But good evaluation work is realized.

(9) Application to other languages In the above-described embodiment, the case where a Japanese input sentence is processed has been described, but an input sentence in another language or an input sentence in which a plurality of languages are mixed are similarly processed. can do. In other words, the “Syntax / semantic analysis result structure (analysis result structure, format)” described above conforms to the structure of “dependency structure” (= word-word dependency = structure showing dependency relationship). It is a concept that is common between languages (the description method and format are also basically common). The mapping rule has the same structure. Therefore, for example, as shown below, when evaluating a skill of an English sentence, if a skill mapping rule is prepared with an English headword, the evaluation can be performed. In addition, if it is specified by meaning instead of headword, a mapping rule common to each language can be described.

(Input sentence)
I developed the accounting system.

(Syntax / semantic analysis result example)
[Develop]
|-(Target case) [accounting system]
|-(Main) [Keyword (I)]

(Example of skill mapping rules)
If <dependency structure> =
[Develop]
|-(Target) [accounting system | application system]
Then <application processing> = QA_SKILL (3) & (answer = Y) & (answer = [heading of the word corresponding to [accounting system | application system])

(10) Others Among the processes described in the above embodiments, all or part of the processes described as being automatically performed can be manually performed, or are described as being manually performed. All or part of the processing can be automatically performed by a known method. In addition, processing procedures, control procedures, specific names, information including various data and parameters (especially the contents of mapping rules, etc.) shown in the above documents and drawings are optional unless otherwise specified. Can be changed.

  Each component of the illustrated devices (particularly, the skill evaluation device 200 illustrated in FIG. 1) is functionally conceptual and does not necessarily need to be physically configured as illustrated. In other words, the specific form of distribution / integration of each device is not limited to that shown in the figure, and all or a part thereof may be functionally or physically distributed or arbitrarily distributed in arbitrary units according to various loads or usage conditions. Can be integrated and configured. Further, all or any part of each processing function performed in each device may be realized by a CPU and a program analyzed and executed by the CPU, or may be realized as hardware by wired logic.

  The processing method according to various processing procedures described in this embodiment (for example, the general-purpose evaluation processing procedure illustrated in FIG. 24) is to execute a program prepared in advance on a computer such as a personal computer or a workstation. Can be realized. This program can be distributed via a network such as the Internet. The program can also be executed by being recorded on a computer-readable recording medium such as a hard disk, a flexible disk (FD), a CD-ROM, an MO, and a DVD and being read from the recording medium by the computer.

  As described above, the evaluation apparatus, the evaluation method, and the evaluation program according to the present invention are useful when extracting and evaluating evaluation items from input sentences described in a natural language, and in particular, in a timely manner in each subcategory. It is suitable for general-purpose evaluation work with good efficiency and quality.

It is a block diagram which shows the structure of the general purpose evaluation apparatus concerning a present Example. It is a figure for demonstrating the outline | summary of the general purpose evaluation process by a present Example. It is a figure for demonstrating the outline | summary of the general purpose evaluation process by a present Example. It is a figure which shows the example of the natural language process by a natural language process part. It is a figure which shows the example of the natural language process by a natural language process part. It is a figure which shows the example of the natural language process by a natural language process part. It is a figure which shows the format of a mapping rule, and the structural example of a skill mapping rule. It is a figure which shows the structural example of a motivation mapping rule, and the structural example of a mental mapping rule. It is a figure for demonstrating the information which can be designated by the condition part of a mapping rule. It is a figure which shows the example of a rule application to a skill question and answer table from a skill mapping rule. It is a figure which shows the example of a rule application to a motivation question and answer table from a motivation mapping rule. It is a figure which shows the example of a rule application from a mental mapping rule to a mental question and answer table. It is a figure which shows the example of a mapping from skill question and an answer table to skill DB. It is a figure which shows the example of a mapping from the motivation question and answer table to motivation DB. It is a figure which shows the example of a mapping from a mental question and an answer table to mental DB. It is a figure which shows the format of a conversion rule, and the structural example of a skill conversion rule. It is a figure which shows the structural example of a motivation conversion rule, and the structural example of a mental conversion rule. It is a figure which shows the example of conversion from skill DB to a skill evaluation table. It is a figure which shows the example of conversion from motivation DB to a motivation evaluation table. It is a figure which shows the example of conversion from motivation DB to a motivation evaluation table. It is a figure which shows the example of conversion from mental DB to a mental evaluation table. It is a figure which shows the format and structural example of the rule for comprehensive evaluation. It is a figure which shows the example of an output of an evaluation result. It is a flowchart which shows the general purpose evaluation process sequence by a present Example. 1 is a system configuration diagram illustrating a configuration of a system including a general-purpose evaluation apparatus according to an embodiment. It is a figure which shows the format of the mapping rule of a word sequence format, the structural example of a skill mapping rule, and the structural example of a motivation mapping rule. It is a figure which shows the example of a direct format mapping. It is a figure which shows the example of a direct format mapping. It is a figure for demonstrating sharing of the condition part in a mapping rule. It is a figure for demonstrating sharing of the condition part in a mapping rule. It is a figure for demonstrating the general purpose evaluation apparatus applied to another category. It is a figure for demonstrating the general purpose evaluation apparatus applied to another category. It is a figure for demonstrating the general purpose evaluation apparatus applied to multiple categories. It is a figure for demonstrating designation | designated of a category. It is a figure for demonstrating the determination of a category. It is a figure for demonstrating the determination of a category.

Explanation of symbols

DESCRIPTION OF SYMBOLS 200 General-purpose evaluation apparatus 201 Natural language process part 202 Mapping rule memory | storage part 202a Skill mapping rule 202b Motivation mapping rule 202c Mental mapping rule 203 Matching part 204 Rule edit part 205 Application process part 206 Question / answer information storage part 206a Skill question / answer table 206b Motivation question / answer table 206c Mental question / answer table 207 Answer complementing processing unit 208 Mapping unit 209 Evaluation result storage unit 209a Skill database (DB)
209b Motivation database (DB)
209c Mental database (DB)
210 Evaluation result analysis unit 211 Evaluation table creation unit 212 Conversion rule storage unit 212a Skill conversion rule 212b Motivation conversion rule 212c Mental conversion rule 212d General evaluation rule

Claims (31)

An evaluation device that extracts and evaluates evaluation items from input sentences described in a natural language,
Evaluation result storage means for storing evaluation results obtained for the evaluation items of each subcategory for each of a plurality of subcategories in a certain category;
For each evaluation item to be extracted in each subcategory, an analysis result assumed as a result of natural language analysis for a sentence including the evaluation item, an evaluation result for the evaluation item, and the evaluation result are stored in the evaluation result memory Mapping rule storage means for storing mapping rules each defining a storage destination when storing in the means;
From the analysis result obtained by performing natural language analysis on the input sentence described in a mixed manner for each subcategory, the evaluation items of each subcategory are extracted using the mapping rule stored in the mapping rule storage unit, and the extracted Mapping processing means for storing an evaluation result corresponding to the evaluation item in a corresponding storage location in the evaluation result storage means;
An evaluation apparatus comprising:   The evaluation apparatus according to claim 1, wherein the mapping rule storage unit stores a mapping rule defined by a common format among the subcategories.   When there is an evaluation item having the same analysis result among a plurality of subcategories, the mapping rule storage unit defines the evaluation result and the storage location for each of the plurality of subcategories in association with the analysis result. The evaluation apparatus according to claim 1, wherein a mapping rule is stored. Conversion rule storage means for storing a conversion rule for converting the evaluation result of each subcategory stored in the evaluation result storage means into a predetermined evaluation table for each subcategory;
Evaluation table creating means for creating an evaluation table for each subcategory from the evaluation result for each subcategory stored in the evaluation result storage means using the conversion rule for each subcategory stored in the conversion rule storage means;
The evaluation apparatus according to claim 1, further comprising: The conversion rule storage means further stores a comprehensive evaluation rule for combining the evaluation results of a plurality of subcategories into a predetermined evaluation table,
The evaluation table creating means creates a comprehensive evaluation table from the evaluation results of a plurality of subcategories stored in the evaluation result storage means, using the comprehensive evaluation rules stored in the conversion rule storage means. The evaluation apparatus according to claim 4. The evaluation result storage means is divided into a plurality of categories and stores the evaluation results of each subcategory for each of a plurality of subcategories belonging to each category,
The mapping rule storage means stores a mapping rule for each evaluation item to be extracted in each subcategory belonging to the plurality of categories,
The said mapping process means extracts and stores using the mapping rule of a corresponding category with respect to the input sentence described mixedly about each subcategory which belongs to one of the categories. The evaluation apparatus according to any one of 5. A receiving means for receiving a category designation for designating which category is an input sentence;
The mapping processing means performs extraction and storage using only the mapping rules of the category relating to the category designation received by the receiving means among the mapping rules of each category stored in the mapping rule storage means. The evaluation apparatus according to claim 6. It further comprises a determination means for determining which category is an input sentence,
The mapping processing unit performs extraction and storage using only the mapping rule of the category determined by the determination unit among the mapping rules of each category stored in the mapping rule storage unit. 6. The evaluation apparatus according to 6.   2. The mapping rule storage unit stores a mapping rule that defines a syntax / semantic structure assumed as a result of syntax / semantic analysis for a sentence including the evaluation item as the analysis result. The evaluation apparatus as described in any one of -8.   9. The mapping rule storage unit stores a mapping rule that defines a word sequence assumed as a result of morphological analysis for a sentence including the evaluation item as the analysis result. The evaluation apparatus as described in any one. Questions / answers for storing a question relating to an evaluation item of each subcategory, an evaluation result obtained as an answer to the question, and a storage location when the evaluation result is stored in the evaluation result storage unit for each of the plurality of subcategories A storage means,
The mapping rule storage means stores a mapping rule that defines a storage location when storing the evaluation result in the question / answer storage means, instead of the storage location of the evaluation result storage means,
The mapping processing means stores an evaluation result corresponding to the evaluation item extracted using the mapping rule in a corresponding storage location in the question / answer storage means, and further stores the evaluation stored in advance in the storage location The evaluation apparatus according to claim 1, wherein the evaluation result is also stored in a storage destination of a result storage unit. For the questions whose evaluation results are not yet stored in the question / answer storage means even after the evaluation items are extracted using the mapping rule, further comprising an answer complementing means for complementing the evaluation results as answers to the questions,
When the evaluation result is complemented by the answer complementing unit, the mapping processing unit stores the evaluation result in a storage destination of the evaluation result storage unit stored in advance in the complementing destination. The evaluation apparatus according to claim 11. Further comprising speech recognition means for recognizing speech information mixed and interviewed for each subcategory to create the input sentence;
The evaluation apparatus according to claim 1, wherein the mapping processing unit extracts and stores the input sentence created by the voice recognition unit. An evaluation method for extracting and evaluating evaluation items from input sentences described in a natural language,
An evaluation result holding step for holding an evaluation result storage means for storing an evaluation result obtained for an evaluation item of each sub category for each of a plurality of sub categories in a certain category;
For each evaluation item to be extracted in each subcategory, an analysis result assumed as a result of natural language analysis for a sentence including the evaluation item, an evaluation result for the evaluation item, and the evaluation result are stored in the evaluation result memory A mapping rule holding step for holding a mapping rule storage means for storing mapping rules each defining a storage destination when storing in the means;
From the analysis result obtained by performing natural language analysis on the input sentence described in a mixed manner for each subcategory, the evaluation items of each subcategory are extracted using the mapping rule stored in the mapping rule storage unit, and the extracted A mapping process for storing an evaluation result corresponding to an evaluation item in a corresponding storage location in the evaluation result storage unit;
The evaluation method characterized by including.   15. The evaluation method according to claim 14, wherein the mapping rule storage means stores a mapping rule defined by a common format among the subcategories.   When there is an evaluation item having the same analysis result among a plurality of subcategories, the mapping rule storage unit defines the evaluation result and the storage location for each of the plurality of subcategories in association with the analysis result. 16. The evaluation method according to claim 14, wherein a mapping rule is stored. A conversion rule holding step for holding a conversion rule storage means for storing a conversion rule for converting the evaluation result of each subcategory stored in the evaluation result storage means into a predetermined evaluation table for each subcategory;
An evaluation table creation step of creating an evaluation table for each subcategory from the evaluation result for each subcategory stored in the evaluation result storage unit, using the conversion rule for each subcategory stored in the conversion rule storage unit;
The evaluation method according to claim 14, 15 or 16, further comprising: The conversion rule storage means further stores a comprehensive evaluation rule for combining the evaluation results of a plurality of subcategories into a predetermined evaluation table,
In the evaluation table creation step, a comprehensive evaluation table is created from the evaluation results of a plurality of subcategories stored in the evaluation result storage unit using the comprehensive evaluation rules stored in the conversion rule storage unit. The evaluation method according to claim 17. The evaluation result storage means is divided into a plurality of categories and stores the evaluation results of each subcategory for each of a plurality of subcategories belonging to each category,
The mapping rule storage means stores a mapping rule for each evaluation item to be extracted in each subcategory belonging to the plurality of categories,
15. The mapping processing step extracts and stores an input sentence described in a mixed manner for each subcategory belonging to any category by using a corresponding category mapping rule. The evaluation method according to any one of 17 above. And further including a receiving step of accepting a category designation for designating which category is an input sentence,
The mapping processing step is to perform extraction and storage using only the mapping rule of the category related to the category designation received by the receiving step among the mapping rules of each category stored in the mapping rule storage unit. The evaluation method according to claim 18. A determination step of determining which category is an input sentence,
The mapping process step performs extraction and storage using only the mapping rule of the category determined by the determination step among the mapping rules of each category stored in the mapping rule storage unit. The evaluation method according to 18. An evaluation program for causing a computer to execute an evaluation method for extracting and evaluating evaluation items from an input sentence described in a natural language,
An evaluation result holding procedure for holding evaluation result storage means for storing the evaluation result obtained for the evaluation item of each subcategory for each of a plurality of subcategories in a certain category;
For each evaluation item to be extracted in each subcategory, an analysis result assumed as a result of natural language analysis for a sentence including the evaluation item, an evaluation result for the evaluation item, and the evaluation result are stored in the evaluation result memory A mapping rule holding procedure for holding a mapping rule storage means for storing mapping rules each defining a storage destination when storing in the means;
From the analysis result obtained by performing natural language analysis on the input sentence described in a mixed manner for each subcategory, the evaluation items of each subcategory are extracted using the mapping rule stored in the mapping rule storage unit, and the extracted A mapping processing procedure for storing an evaluation result corresponding to an evaluation item in a corresponding storage destination in the evaluation result storage unit;
An evaluation program for causing a computer to execute.   The evaluation program according to claim 22, wherein the mapping rule storage unit stores a mapping rule defined by a common format among the subcategories.   When there is an evaluation item having the same analysis result among a plurality of subcategories, the mapping rule storage unit defines the evaluation result and the storage location for each of the plurality of subcategories in association with the analysis result. 24. The evaluation program according to claim 22 or 23, wherein a mapping rule is stored. A conversion rule holding procedure for holding conversion rule storage means for storing a conversion rule for converting the evaluation result of each subcategory stored in the evaluation result storage means into a predetermined evaluation table for each subcategory;
An evaluation table creation procedure for creating an evaluation table for each subcategory from the evaluation result for each subcategory stored in the evaluation result storage unit, using the conversion rule for each subcategory stored in the conversion rule storage unit;
25. The evaluation program according to claim 22, 23 or 24, further causing a computer to execute. The conversion rule storage means further stores a comprehensive evaluation rule for combining the evaluation results of a plurality of subcategories into a predetermined evaluation table,
In the evaluation table creation procedure, a comprehensive evaluation table is created from the evaluation results of a plurality of subcategories stored in the evaluation result storage unit using the comprehensive evaluation rules stored in the conversion rule storage unit. The evaluation program according to claim 25. The evaluation result storage means is divided into a plurality of categories and stores the evaluation results of each subcategory for each of a plurality of subcategories belonging to each category,
The mapping rule storage means stores a mapping rule for each evaluation item to be extracted in each subcategory belonging to the plurality of categories,
23. The mapping processing procedure includes: extracting and storing an input sentence described in a mixed manner with respect to each subcategory belonging to any category by using a corresponding category mapping rule. 26. The evaluation program according to any one of 26. Let the computer further execute a reception procedure for accepting a category designation that specifies which category the input sentence relates to,
The mapping processing procedure is characterized in that extraction and storage are performed using only the category mapping rule relating to the category designation accepted by the acceptance procedure, out of the mapping rules of each category stored in the mapping rule storage means. The evaluation program according to claim 27. Let the computer further execute a determination procedure for determining which category the input sentence is,
The mapping processing procedure is to perform extraction and storage using only the mapping rule of the category determined by the determination procedure among the mapping rules of each category stored in the mapping rule storage unit. 27. The evaluation program according to 27. An evaluation device that extracts and evaluates evaluation items from input sentences described in a natural language,
Evaluation result storage means for storing evaluation results obtained for the evaluation items of each category for each of a plurality of categories;
For each evaluation item to be extracted in each category, an analysis result assumed as a result of natural language analysis for a sentence including the evaluation item, an evaluation result for the evaluation item, and the evaluation result are stored in the evaluation result memory Mapping rule storage means for storing mapping rules each defining a storage destination when storing in the means;
From the analysis result obtained by performing natural language analysis on the input sentence described in a mixed manner for each category, the evaluation items of each category are extracted using the mapping rule stored in the mapping rule storage unit, and the extracted Mapping processing means for storing an evaluation result corresponding to the evaluation item in a corresponding storage location in the evaluation result storage means;
An evaluation apparatus comprising: An evaluation program for causing a computer to execute an evaluation method for extracting and evaluating evaluation items from an input sentence described in a natural language,
An evaluation result holding procedure for holding an evaluation result storage means for storing an evaluation result obtained for an evaluation item of each category for each of a plurality of categories;
For each evaluation item to be extracted in each category, an analysis result assumed as a result of natural language analysis for a sentence including the evaluation item, an evaluation result for the evaluation item, and the evaluation result are stored in the evaluation result memory A mapping rule holding procedure for holding a mapping rule storage means for storing mapping rules each defining a storage destination when storing in the means;
From the analysis result obtained by performing natural language analysis on the input sentence described in a mixed manner for each category, the evaluation items of each category are extracted using the mapping rule stored in the mapping rule storage unit, and the extracted A mapping processing procedure for storing an evaluation result corresponding to an evaluation item in a corresponding storage destination in the evaluation result storage unit;
An evaluation program for causing a computer to execute.

Images