JP5265597B2 - Document quality evaluation system and document quality evaluation program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a document quality evaluation system that presents examples of a sentence to be improved and a sample sentence, in addition to a result of evaluating quality of a text document group, to an input person. <P>SOLUTION: The document quality evaluation system 1 for evaluating the quality of the text document group and presenting the evaluation result to a user includes: a parsing part 10 parsing each sentence in the text document group; a quality evaluation part 20 scoring the quality of the text document group for a plurality of evaluation items, and presenting the evaluation result to a first user; and an example sentence extraction part 30 extracting a first sentence group corresponding to a low rank in evaluation in a lowest evaluation item for the first user and a second sentence group corresponding to a condition of high rank in evaluation for a second user ranked higher than the first user in the lowest evaluation item, extracting first and second sentences having high similarity from the first and second sentence groups, presenting the first sentence as the example sentence for improvement, and presenting the second sentence as the example sentence for a sample. <P>COPYRIGHT: (C)2011,JPO&amp;INPIT

Description

  The present invention relates to a language processing technique, and more particularly to a technique effective when applied to a document quality evaluation system and a document quality evaluation program for evaluating the quality of a text document in terms of syntax analysis.

  For example, in companies that provide services to customers, the contents of questions, consultations, complaints, etc. received from customers by telephone or e-mail etc. by operators at call centers etc. are recorded as text documents, and the text documents stored later It is performed to obtain various knowledge by analyzing the text by text mining. At this time, if the contents such as the description method of the text input by the operator or the like are not uniform, it is difficult to improve the accuracy of analysis for these. Therefore, it is necessary to ensure a certain quality for the input text document.

  In order to ensure the quality of the input text document, in addition to support for education to the operator, etc., as support by the information processing system, for example, sentences in the input text document are analyzed by syntax analysis, etc. For example, the quality of the text document (group) is evaluated by scoring according to the above conditions, and an evaluation result and an improvement point are presented.

  Further, for example, in Japanese Patent Application Laid-Open No. 2007-164673 (Patent Document 1), in creating a product manual or the like, a language that is a target of document check in order to improve the quality and efficiency of a document created by a user. Control the execution of the rule check, the execution of the background information check, and the output of the check result based on the association information that associates the rule whose expression is described in the predetermined format with the background information described in the predetermined format. Enables execution of background information check based on the result of rule check, execution of rule check based on the result of background information check, browsing of background information associated with the result of rule check, etc., enabling effective document creation support by the user Techniques that make it possible are described.

JP 2007-164673 A

  For example, unlike the call log at the call center or the like, the technique described in Patent Document 1 targets a product manual or the like, and therefore it is necessary to reliably exclude inappropriate descriptions. Must be extracted and presented to the user. However, unlike the preparation of product manuals, operators such as call centers need to input a variety of contents in a concise and speedy manner. Is big. Also, since the input content has a high degree of freedom, it is difficult to create rules and templates.

  On the other hand, in the method of evaluating and presenting input text documents (groups) by scoring etc., the overall quality level can be grasped, but what kind of sentences and descriptions are specifically evaluated? It is difficult for a user such as an operator to accurately grasp information such as how the information should have been described, and the quality of input by the user may not be improved.

  Therefore, an object of the present invention is to provide an example of a text to be improved and an example of a text to be used as a sample in addition to the evaluation result of the quality of the input text document group to a text document input person. An object of the present invention is to provide a document quality evaluation system and a document quality evaluation program that can be presented as information. The above and other objects and novel features of the present invention will be apparent from the description of this specification and the accompanying drawings.

  Of the inventions disclosed in this application, the outline of typical ones will be briefly described as follows.

  A document quality evaluation system and a document quality evaluation program according to exemplary embodiments of the present invention take in a text document group input by a user, and the target text document group for each of the user or a group to which the user belongs. The quality of the text document is evaluated by scoring from the viewpoint of syntax analysis, and the evaluation result is presented to the user, thereby supporting the user to improve the input quality of the text document. Have the following characteristics.

  That is, the document quality evaluation system and the document quality evaluation program perform morphological analysis and syntax analysis on each sentence in the captured text document group, perform aggregation processing based on the analysis result, and calculate the analysis result and the aggregation result. Based on the information stored in the table and the information held in the table, the quality of the target text document group is parsed for each first user or each group to which the first user belongs. A quality evaluation unit that scores and evaluates a plurality of evaluation items from a viewpoint, outputs an evaluation result, and presents the result to the first user.

  Furthermore, the first sentence group of the first user corresponding to a condition of low evaluation in the lowest evaluation item that is the lowest evaluation item in the evaluation result for the first user, and the lowest evaluation A second sentence group corresponding to a condition that is high in the lowest evaluation item for a second user whose evaluation is higher than that of the first user in the item, and the first sentence group and the A first sentence and a second sentence having a high degree of similarity are extracted from the second sentence group, the first sentence is used as an example sentence to be improved, and the second sentence is output as an example sentence. And an example sentence extraction unit to be presented to the first user.

  Among the inventions disclosed in the present application, effects obtained by typical ones will be briefly described as follows.

  According to the exemplary embodiment of the present invention, the text document group input by the user is evaluated by scoring the quality for a plurality of evaluation items from the viewpoint of syntax analysis, and for the evaluation item with the lowest score, Examples of sentences that should be improved and examples that should be sampled by extracting and presenting similar sentence pairs from sentence groups that meet the low evaluation conditions and sentence groups that meet the high evaluation conditions Can be presented.

  As a result, it is possible to provide support for the user to improve the input quality of the text in a concise, specific and accurate manner. Furthermore, since the quality of the stored text document is improved, analysis by text mining or the like can be performed effectively.

It is the figure which showed the outline | summary of the structural example of the document quality evaluation system which is one embodiment of this invention. It is the figure shown about the data structure of the text document management table in one embodiment of this invention, and the example of concrete data. It is the figure shown about the data structure and specific example of data of the word management in one embodiment of this invention. It is the figure shown about the data structure of the text management in one embodiment of this invention, and the example of concrete data. It is the flowchart which showed the example of the process which extracts the improvement object and sample example sentence in one embodiment of this invention. It is the figure which showed the example in the case of calculating the cosine similarity between sentences in one embodiment of this invention.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. Note that components having the same function are denoted by the same reference symbols throughout the drawings for describing the embodiment, and the repetitive description thereof will be omitted.

  The document quality evaluation system according to an embodiment of the present invention includes, for example, a call log (response history) in a call center system or the like for contents such as questions, consultations, and complaints received from customers by telephone or e-mail. ), The quality of the target text document group is scored and evaluated from the viewpoint of parsing for each user (operator, etc.) and group, and the evaluation result is presented to the user. Is a computer system having a function of providing support for improving the input quality of a text document.

  When evaluating the quality, evaluation is performed using five evaluation items described later from the viewpoint of syntax analysis, and each evaluation item is scored and statistical information is calculated and output. In addition, a sentence group corresponding to a condition that is low in the evaluation item with the lowest score in the target user and a sentence group that corresponds to a condition that is high in the evaluation item including other users are extracted, A pair of sentences having a high similarity is extracted from these sentence groups and presented to the user as an example sentence to be improved and an example sentence of the sample.

[System configuration]
FIG. 1 is a diagram showing an outline of a configuration example of a document quality evaluation system according to an embodiment of the present invention. The document quality evaluation system 1 is constituted by, for example, a server, a PC (Personal Computer), etc., and each unit of a syntax analysis unit 10, a quality evaluation unit 20, an example sentence extraction unit 30 and an interface unit 40 implemented by a software program, and a database And text document management 51, word management 52, and sentence management 53 including file tables and the like. Further, it has a redundant element 11 composed of a file table or the like.

  The syntax analysis unit 10 performs so-called morphological analysis and syntax analysis on each sentence in the text document captured via the interface unit 40 described later, and performs analysis processing such as decomposition into words and generation of a syntax tree. It has a function. For each analysis process of morpheme analysis and syntax analysis, known techniques and algorithms in natural language processing can be used as appropriate, and a general morpheme analysis engine or syntax analysis engine is incorporated into the syntax analysis unit 10 for use. Is also possible.

  The results of the analysis processing are stored in the text document management 51, word management 52, and sentence management 53 tables. At this time, based on the analysis result, the number of words in each sentence, the number of expressions judged to be a situation description, and whether the sentence has a subject to be used for evaluation in the quality evaluation unit 20 described later Whether or not the table of the redundant element 11 has a redundant word such as a greeting set in advance is performed, and the information is also stored in the table of the sentence management 53.

  Based on the information held in the text document management 51, word management 52, and sentence management 53 tables, the quality evaluation unit 20 parses the quality of the target text document group for each user or group. It has a function of scoring and evaluating five evaluation items to be described later from the viewpoint, and presenting the evaluation result to the user via the interface unit 40. Further, it may have a function of calculating and presenting statistical information such as various averages, ratios, and deviation values for the evaluation result, and a function of storing and holding the evaluation result in a database or the like.

  The example sentence extraction unit 30 includes a sentence group (which is input by the target user) corresponding to a condition in which the evaluation item having the lowest evaluation (score) in the target user is low, and a condition in which the evaluation item is high in evaluation. As an example sentence to be improved and a sample sentence, a sentence group corresponding to the above (including those input by the target user and other users) is extracted, and a sentence pair having a high similarity is extracted from these sentence groups. The function of presenting to the user via the interface unit 40 is provided. You may have the function to store and hold the extraction result of the example of improvement and a sample sentence in a database etc.

  The interface unit 40 has a user interface such as screen display in the document quality evaluation system 1 and an input / output function such as capturing a text document group. The display of the screen or the like is not shown, but may be displayed directly on a display or the like provided in the document quality evaluation system 1 itself, or via a Web browser on the client terminal using a Web server program. A screen may be displayed. In the present embodiment, the document quality evaluation system 1 is implemented as an independent system. However, for example, it is implemented as a part of another system such as a call center system that accepts input of a text document from a user. May be.

[Data structure]
FIG. 2 is a diagram showing a data structure of the text document management 51 and an example of specific data. The text document management 51 is a table for managing a text document composed of sentences input by a user, and has items such as a text document ID and an input person attribute, for example. The text document ID item holds ID information that is assigned by the document quality evaluation system 1 and uniquely identifies each text document. The item of the input person attribute holds information on the attribute of the user (input person) who inputs the target text document, and holds at least information such as an ID and a user name that can uniquely identify the user.

  FIG. 3 is a diagram illustrating a data configuration of the word management 52 and a specific example of data. The word management 52 is a table for managing each word included in the sentence input by the user, and includes items such as a text document ID, a sentence number, and word data. The text document ID item holds information on the ID of the text document to which the sentence including the target word belongs. This text document ID corresponds to the text document ID item of the text document management 51 in FIG.

  The item of the sentence number holds an identification number assigned by the document quality evaluation system 1 for the sentence including the target word in the target text document. This sentence number corresponds to a sentence number item in the sentence management 53 of FIG. The item of word data holds text data of a word obtained as a result of the target sentence analysis processing by the syntax analysis unit 10.

  FIG. 4 is a diagram showing a data configuration of the sentence management 53 and an example of specific data. The sentence management 53 is a table for managing each sentence input by the user, and includes items such as a text document ID, a sentence number, sentence data, a word count, a situation description count, a subject flag, and a redundant element flag. The text document ID item holds information on the ID of the text document to which the target sentence belongs. This text document ID corresponds to the text document ID item of the text document management 51 in FIG.

  The item of sentence number holds information on the identification number assigned by the document quality evaluation system 1 in the target text document for the target sentence. One text document includes one or more sentences, and the identification number value of each sentence is assigned, for example, 1, 2,... In order from the first sentence in the text document. The text data item holds text data of the target text.

  The word count item holds information on the number of words included in the target sentence. The number of situation description holds information on the number of expressions determined to be the situation description in the target sentence. The subject flag holds flag value information indicating whether or not the subject sentence has a subject. The redundant element flag holds flag value information indicating whether or not the target sentence has redundant words such as greetings preset in the table of the redundant element 11. The information of each item is information obtained based on the result of the analysis process for the target sentence by the syntax analysis unit 10.

  The above tables and their data structures are not limited to those shown in the drawings, and may be other table structures or have other items as long as the data items described above can be managed. Good.

[Scoring]
When the parsing unit 10 performs analysis processing on the captured text document group and information based on the analysis result is held in the text document management 51, word management 52, and sentence management 53 tables, quality evaluation is performed next. The unit 20 evaluates the quality of the input text document group for each user or group from the viewpoint of syntax analysis. In the present embodiment, it is an attention point of a sentence in a call log in a call center and can be easily evaluated from the result of syntax analysis. For example, (1) “sentence of sentence”, (2) “main predicate relation” , (3) “Explanation element”, (4) “Situation description”, (5) “Exclusion redundancy” is scored and evaluated. These evaluation items can be other evaluation items according to the characteristics of the text document to be evaluated.

  (1) In the evaluation of “sentence of sentence”, scoring is performed on the condition that the sentence in the text document is established as a sentence rather than a description of only words. The score S for “sentence of sentence” is calculated, for example, as the ratio of all text documents excluding text documents consisting only of sentences that have no room for establishment as sentences, with respect to all text documents.

  Here, the number of text documents consisting only of sentences that have no room for a sentence is, for example, a text document that does not have a sentence consisting of two or more words (a text document consisting only of sentences having only one word). As the number of Note that the number of words included in the sentence can be grasped from the word number item of the sentence management 53. Further, whether or not the target text document corresponds to the target user or group can be determined based on the item of the input person attribute of the text management 51.

  (2) In the evaluation of “main predicate relationship”, scoring is performed on the condition that the sentence in the text document is a sentence having the main predicate relation of the subject / predicate. The score S for the “subject relationship” is, for example, the ratio of all text documents excluding text that has no room for establishment as the above-mentioned sentence and text documents consisting only of sentences having no subject, It is calculated by the following formula.

  Note that a sentence having no subject can be recognized as a sentence whose subject flag item of the sentence management 53 is FALSE, for example.

  (3) In the evaluation of “explanatory element”, scoring is performed on the condition that the sentence in the text document has an explanatory element other than the subject / predicate. The score S for the “explanatory element” excludes, for example, a text document that has no room for formation as a sentence described above from all text documents, a sentence that does not have a subject, and a text document that includes only a sentence composed only of the subject and a predicate. As a percentage of all text documents, the following formula is used.

  In addition, about the sentence comprised only with a subject and a predicate, it can be grasped | ascertained as a sentence in which the item of the word count of the sentence management 53 is less than 3 among the sentences which have a subject relation.

  (4) In the evaluation of “situation description”, scoring is performed on the condition that the sentence in the text document has a description about the situation. The score S for “situation description” is calculated by, for example, the following formula as a ratio of a text document including a sentence having a subject and a description about the situation to all text documents.

  Note that a sentence having a description about a situation can be understood as a sentence in which the item of the number of situation description in the sentence management 53 is greater than zero. Moreover, these sentences do not correspond to the sentence excluded in evaluation of said (1)-(3).

  (5) In the evaluation of “redundancy redundancy”, scoring is performed on the condition that a sentence in a text document does not have a redundant word such as a greeting. The score S for “redundancy redundancy” is calculated, for example, as the ratio of all text documents excluding text documents having sentences with redundant words to all text documents by the following formula.

  A sentence having redundant words can be recognized as a sentence whose redundant element flag item in the sentence management 53 is TRUE.

  The score S in each evaluation item for the input text document is calculated for each user or each group by the equations shown in the above formulas 1 to 5, and presented to the user via the interface unit 40. The form of presentation is not particularly limited, and the score may be presented as it is for each evaluation item, or it is presented after being converted into a rank (for example, “A” to “E”, etc.) predefined by the range of the score. Also good. At this time, depending on the group configuration, for example, all users fall under the rank “A”, and considering the case where accurate evaluation (particularly pointing out improvement points) is difficult, the weighting value for scoring is set. It may be possible to set the weight (for example, in the case of setting for “strict” evaluation, the weighting value is set so that the value of the score S becomes small).

[Example sentence extraction]
After the score (rank) for each evaluation item for the text document group is calculated for each user or for each group by the quality evaluation unit 20, the example sentence extraction unit 30 provides an example sentence for improvement to be presented to the target user and Extract sample sentences (important points in the text document). FIG. 5 is a flowchart illustrating an example of processing for extracting an improvement target and a sample example sentence.

  First, the information of the evaluation result about each evaluation item evaluated by the quality evaluation part 20 is acquired about the target user (input person 1) (S101). Next, it is determined whether or not each evaluation item has the highest rank “A” (or a score equal to or higher than a predetermined numerical value) (S102). If all the evaluation items are ranked “A”, a message indicating that there is no improvement point is output via the interface unit 40, for example, “There is no item to be improved” (S110), and the processing is performed. finish.

  If all the evaluation items are not ranked “A” in step S102, the evaluation item with the lowest rank (or score) is acquired (S103). Next, the sentence of the input person 1 corresponding to the condition of low evaluation in the evaluation item with the lowest rank is extracted as the sentence group A, and this information is stored in the array A (S104). Specifically, the text document input by the input person 1 is specified by the text management 51, and the text document ID and the sentence of each sentence corresponding to the condition that the sentence management 53 receives a low evaluation from the sentence included in the target text document. The number item is extracted, and this value is stored in the array A.

Here, the conditions for low evaluation in each evaluation item correspond to the conditions for scoring in the quality evaluation unit 20 described above,
“Sentence formation”: The number of words item in the sentence management 53 is 1 or less. “Subject relationship”: The subject flag item in the sentence management 53 is FALSE. “Description element”: Item in the number of words in the sentence management 53 Is less than 3 “Situation Description”: The item of the number of situation description in the sentence management 53 is 0 “Exclusion Redundancy”: A sentence group corresponding to each condition that the redundancy element flag item in the sentence management 53 is TRUE Will be extracted.

  Next, it is determined whether or not there is a user (input person 2) whose rank (or score) is higher than that of the input person 1 in the target evaluation item (an evaluation item having the lowest rank (or score) for the input person 1). (S105). If there is an input person 2 having a high evaluation, the sentence of the input person 2 corresponding to the condition of high evaluation in the target evaluation item is extracted as a sentence group B, and this information is stored in the array B (S107). The process proceeds to step S109. Specifically, the text document input by the input person 2 is identified by the text management 51, and the text document ID and the sentence of each sentence corresponding to a condition that is highly evaluated from the sentences included in the target text document in the sentence management 53 The number item is extracted, and this value is stored in the array B.

Here, the conditions for high evaluation in each evaluation item correspond to the conditions for scoring in the quality evaluation unit 20 described above (the reverse of the conditions for low evaluation described above),
“Sentence formation”: The number of words in the sentence management 53 is 2 or more “Subject relationship”: The subject flag item in the sentence management 53 is TRUE “Description element”: The number of words in the sentence management 53 Is “3” or more. “Situation Description”: The item of the number of situation description in the sentence management 53 is 1 or more. “Exclusion Redundancy”: The sentence corresponding to each condition that the redundant element flag item in the sentence management 53 is FALSE. A group will be extracted.

  If there is no input person 2 having a high evaluation in step S105, it is determined whether or not there is a sentence of the input person 1 corresponding to a high evaluation condition in the target evaluation item (S106). If there is no corresponding sentence of the input person 1, an example sentence to be improved for the input person 1 is extracted from the information of the sentence group A stored in the array A and output via the interface unit 40 (S 113). Specifically, for example, data with the smallest (smallest) text document ID and sentence number values is selected from the array A, and sentence data specified by these values is acquired from the sentence management 53 to be improved. And Further, through the interface unit 40, a message indicating that there is no sample sentence as a sample, such as “There is no sample sentence” is output (S114), and the process is terminated.

  If there is a sentence of the input person 1 corresponding to the high evaluation condition in step S106, the sentence of the input person 1 corresponding to the high evaluation condition in the target evaluation item is extracted as the sentence group B, and this information is extracted. Store in array B (S108). Since the specific process of extracting the sentence group B is the same as that in step S107 described above, description thereof is omitted.

  Next, the sentence pair (extracted from the sentence a and the sentence group B extracted from the sentence group A) having the highest similarity in the sentence group A and the sentence group B (extracted in step S107 or S108). Sentence b) is extracted (S109). When there are a plurality of sentences having the highest similarity in the sentence group A or the sentence group B, for example, the text document ID and sentence number values in the array A or the array B are the youngest (smallest) data. And the sentence specified by these values is designated as sentence a or sentence b.

  Various methods can be used for calculating the similarity between the sentences. For example, a frequency vector for each sentence is calculated from the appearance frequency of words included in each sentence obtained based on the word management 52, and a so-called cosine similarity is calculated from the angle formed between the frequency vectors to obtain the similarity. it can.

  FIG. 6 is a diagram illustrating an example of calculating the cosine similarity between sentences. In FIG. 6, the sentence d1 includes the words t1, t2, t3, and t5, and the sentence d2 indicates that the words t1, t2, t4, and t5 are included. Here, the frequency vectors U1 and U2 of the sentences d1 and d2 are determined so as to illustrate the appearance frequencies of the words t1 to t5 in the sentences d1 and d2 as elements. Here, the cosine similarity cos (d1, d2) between the sentences d1 and d2 can be obtained by dividing the inner product of the frequency vectors U1 and U2 by the magnitude of each frequency vector, as shown in the figure.

  After that, among the extracted sentence pairs, the sentence a (sentence input by the input person 1) is output as an example sentence to be improved for the input person 1 via the interface unit 40 (S111), and further the sentence b (input person 2) Alternatively, the sentence entered by the input person 1 is output as an example sentence via the interface unit 40 (S112), and the process is terminated. Since the sentence a and sentence b of the sentence pair have a high degree of content similarity, the example sentence to be improved and the example sentence to be sampled are similar in contents. Therefore, the user can accurately grasp how to describe the example sentence to be improved by referring to the example sentence as a sample.

  When presenting an example sentence to be improved and a sample to the user via the interface unit 40, for example, the evaluation item having the lowest rank is pointed out together with a message or the like, and the example sentence to be improved about the evaluation item and Sample example sentences are output side by side. For the example sentence as a sample, attributes such as the name of the user (input person 2 or input person 1) who input the example sentence, and rank (score) information about the evaluation item may be output together. .

  As described above, according to the document quality evaluation system 1 according to an embodiment of the present invention, the text document group input by the user is scored with respect to a plurality of evaluation items from the viewpoint of parsing. In addition to evaluating, for the evaluation item with the lowest score, by extracting and presenting a pair of sentences with a high degree of similarity from a sentence group corresponding to the low evaluation condition and a sentence group corresponding to the high evaluation condition, It is possible to present an example sentence to be improved and an example sentence.

  As a result, it is possible to provide support to the user by presenting information for improving the text input quality in a concise, specific and accurate manner. Furthermore, since the quality of the stored text document is improved, analysis by text mining or the like can be performed effectively.

  As mentioned above, the invention made by the present inventor has been specifically described based on the embodiment. However, the present invention is not limited to the embodiment, and various modifications can be made without departing from the scope of the invention. Needless to say.

  The present invention is applicable to a document quality evaluation system and a document quality evaluation program for evaluating the quality of a text document from the viewpoint of syntax analysis.

1 ... Document quality evaluation system,
DESCRIPTION OF SYMBOLS 10 ... Syntax analysis part, 11 ... Redundant element, 20 ... Quality evaluation part, 30 ... Example sentence extraction part, 40 ... Interface part,
51 ... Text document management, 52 ... Word management, 53 ... Sentence management.

Claims (10)

A text document group input by a user is captured, and for each user or group to which the user belongs, the quality of the target text document group is scored and evaluated from the viewpoint of syntax analysis, and the evaluation result is given to the user. A document quality evaluation system that provides support for improving the quality of input of the text document to the user by presenting,
A parse analysis unit that performs morphological analysis and syntax analysis for each sentence in the captured text document group, performs aggregation processing based on the analysis result, and stores the analysis result and the aggregation result in a table;
Based on the information held in the table, for each of the first user or the group to which the first user belongs, the quality of the target text document group is scored for a plurality of evaluation items from the viewpoint of parsing And a quality evaluation unit that outputs an evaluation result and presents it to the first user;
In the first sentence group of the first user corresponding to the condition of low evaluation in the lowest evaluation item that is the lowest evaluation item in the evaluation result for the first user, and in the lowest evaluation item A second sentence group corresponding to a condition in which the second user having a higher evaluation than the first user is highly evaluated in the lowest evaluation item is extracted, and the first sentence group and the second sentence group are extracted. The first sentence and the second sentence having high similarity are extracted from the sentence group, the first sentence is used as an example sentence to be improved, and the second sentence is output as an example sentence. A document quality evaluation system comprising an example sentence extraction unit to be presented to one user. The document quality evaluation system according to claim 1,
When the second user whose evaluation is higher than that of the first user in the lowest evaluation item does not exist in the lowest evaluation item, the example sentence extraction unit extracts the first user who satisfies the condition of high evaluation in the lowest evaluation item. A document quality evaluation system, wherein a sentence group is extracted and used as the second sentence group. The document quality evaluation system according to claim 2,
The example sentence extraction unit, when there is no sentence of the first user that meets the condition of high evaluation in the lowest evaluation item, the sentence selected from the first sentence group is set as the first sentence, A document quality evaluation system characterized in that a first sentence is used as the example sentence to be improved, and the fact that there is no corresponding example sentence is output and presented to the first user. In the document quality evaluation system according to any one of claims 1 to 3,
The example sentence extracting unit outputs that there is no improvement point and presents it to the first user when all the evaluation results of the respective evaluation items for the first user are evaluations equal to or higher than a predetermined value. Document quality evaluation system characterized by In the document quality evaluation system according to any one of claims 1 to 4,
The document example evaluation unit uses a cosine similarity based on a frequency vector having an appearance frequency of each word included in each target sentence as the similarity. A text document group input by a user is captured, and for each user or group to which the user belongs, the quality of the target text document group is scored and evaluated from the viewpoint of syntax analysis, and the evaluation result is given to the user. A document quality evaluation program that causes a computer to function as a system for providing support for improving the quality of input of the text document to the user by presenting,
A parse analysis unit that performs morphological analysis and syntax analysis for each sentence in the captured text document group, performs aggregation processing based on the analysis result, and stores the analysis result and the aggregation result in a table;
Based on the information held in the table, for each of the first user or the group to which the first user belongs, the quality of the target text document group is scored for a plurality of evaluation items from the viewpoint of parsing And a quality evaluation unit that outputs an evaluation result and presents it to the first user;
In the first sentence group of the first user corresponding to the condition of low evaluation in the lowest evaluation item that is the lowest evaluation item in the evaluation result for the first user, and in the lowest evaluation item A second sentence group corresponding to a condition in which the second user having a higher evaluation than the first user is highly evaluated in the lowest evaluation item is extracted, and the first sentence group and the second sentence group are extracted. The first sentence and the second sentence having high similarity are extracted from the sentence group, the first sentence is used as an example sentence to be improved, and the second sentence is output as an example sentence. A document quality evaluation program comprising an example sentence extraction unit presented to one user. The document quality evaluation program according to claim 6,
When the second user whose evaluation is higher than that of the first user in the lowest evaluation item does not exist in the lowest evaluation item, the example sentence extraction unit extracts the first user who satisfies the condition of high evaluation in the lowest evaluation item. A document quality evaluation program, wherein a sentence group is extracted and used as the second sentence group. The document quality evaluation program according to claim 7,
The example sentence extraction unit, when there is no sentence of the first user that meets the condition of high evaluation in the lowest evaluation item, the sentence selected from the first sentence group is set as the first sentence, A document quality evaluation program characterized in that a first sentence is used as the example sentence to be improved and a message indicating that there is no corresponding example sentence is output and presented to the first user. In the document quality evaluation program according to any one of claims 6 to 8,
The example sentence extracting unit outputs that there is no improvement point and presents it to the first user when all the evaluation results of the respective evaluation items for the first user are evaluations equal to or higher than a predetermined value. A document quality evaluation program characterized by The document quality evaluation program according to any one of claims 6 to 9,
The example sentence extraction unit uses a cosine similarity based on a frequency vector having an appearance frequency of each word included in each target sentence as the similarity.

Images