JP2002140465A - Natural sentence processor and natural sentence processing program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To accurately acquire opinions from free description type questionnaire answer sentences or the like without any human labor. SOLUTION: This natural sentence processor for analyzing opinions provided by a plurality of opinion providers A is provided with an answer extracting means 2 for extracting the electronic data of the free description parts of the opinion providers from the provided opinions as the answer sentences of opinion provision, an opinion extracting means 3 for extracting pairs of words from the extracted answer sentences, an opinion totaling means 4 for totaling the respective extracted opinion contents, and a result providing means 5 for providing the totaled result of the respective opinion contents.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a natural sentence processing device and a natural sentence processing program written in free description format.
For example, the present invention relates to a document processing apparatus and a document processing program for analyzing, as a result of a questionnaire, a response to a questionnaire performed in a market research or the like when the answer is a natural sentence.

[0002]

2. Description of the Related Art A questionnaire is used as one of means for conducting various surveys. For the questionnaire, a free-form questionnaire system in which some or all of the answers are free-form documents, and a questionnaire questionnaire with several options prepared in advance, and the questionnaire respondents can select from among them There is a selection method to make it.

[0003] The open-ended questionnaire method has a problem that the answer result is a natural sentence, and manual counting and analysis are required to obtain useful information from many respondents, which is very costly. is there.

[0004] As a means for solving this problem, there is the above-mentioned selection method. However, in the case of this method, there is a problem in that information regarding options other than those created in advance cannot be obtained, and important knowledge such as needs that the questionnaire implementer did not expect cannot be obtained.

[0005] A questionnaire totalizing system has been proposed (JP-A-09-190454) for automatically totaling by a computer at the time of totaling in order to reduce the cost of totalizing the free-form questionnaire.

[0006] This questionnaire totaling system is a device that has a thesaurus dictionary in the device, searches the thesaurus dictionary for each word in a free description sentence, and integrates the results by its higher terms. For example, it is a device that performs processing such as summing up “hardball baseball” and “softball baseball” as an upper word “baseball”.

[0007] However, since this integration is performed on a word-by-word basis, as a response to the question item "contents of lifelong learning" of the questionnaire mentioned in the description of this device, for example, "hardball baseball" and "softball baseball.""." Can be correctly counted as a "baseball" sentence that is described only in words, but "I thought that when I was young, I thought baseball, but now golf is still better. If it is described as a negative object such as "I think", there is a problem that accurate counting cannot be performed.

There are also questions that this method makes accurate summarization almost impossible. For example, regarding a question such as "Please give us your opinion about this product", the questionnaire writer does not answer with only words, and also asks whether the description is negative or positive. For the purpose, it is necessary to make a distinction at the time of counting, but this cannot be done with this device.

In other words, in this device, a free-form questionnaire in the original sense is permitted, that is, any question sentence is permitted as a questionnaire performer, and any description is permitted as a questionnaire description side. Cannot be calculated accurately.

In addition to the questionnaire, examples in which a large number of people express their opinions include articles in bulletin boards on personal computer communications and web pages, and customer inquiries sent to help desks. The opinion of each individual is described in a natural sentence, and there is a problem that it takes the same cost as the analysis work of the open-ended questionnaire to obtain useful information from them.

[0011]

SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to solve the above-described problems in the free-form questionnaire and the like, and to obtain useful information accurately at low cost.

Further, the present invention is not limited to a free description type questionnaire, but is intended to obtain useful information accurately and at low cost from an information source in which each individual's opinion is described in the form of a natural sentence. .

[0013]

FIG. 1 shows a principle configuration for achieving the object of the present invention. In FIG. 1, 1 is an input unit, 2 is an answer extracting unit, 3 is an opinion extracting unit, 4 is an opinion totaling unit, 5 is a result presenting unit, 6 is an opinion determining unit, 7 is a judgment knowledge acquiring unit, and 8 is totaling knowledge. Reference numeral 9 denotes a judgment knowledge database, and reference numeral 10 denotes a tally result storage database.

The above object of the present invention can be achieved by the following constitutions (1) to (7).

(1) In a natural sentence processing apparatus that analyzes opinions provided by a plurality of opinion providers, digitized data of a free description portion of the opinion provider is extracted from the provided opinions as a reply sentence for providing an opinion. Answer extracting means 2, Opinion extracting means 3 for extracting word pairs from the extracted answer sentences, Opinion totaling means 4 for totalizing each extracted opinion content, and providing results for providing the totaling result for each opinion content It is characterized by comprising means 5.

(2) In the natural sentence processing apparatus according to (1), opinion judging means 6 for judging whether the extracted opinion gives a positive evaluation or a negative evaluation.
And a judgment knowledge database 9 for storing knowledge for judging opinions.

(3) In the natural sentence processing apparatus according to the above (2), when two positive opinions and negative opinions are obtained in one answer sentence, the strength is determined and the overall answer sentence is determined. It is characterized by having opinion judgment means 6 for judging positive or negative.

(4) The natural sentence processing apparatus according to (1) or (2), further comprising a result presenting means 5 for displaying the presented information as a plurality of graphs in which the whole set is changed.

(5) In the natural sentence processing apparatus according to the above (2) or (3), knowledge for judging an extracted opinion is automatically acquired from an answer result in an opinion survey form of a specific format. It is characterized by comprising judgment knowledge acquiring means 7-1 for accumulating.

(6) In the natural sentence processing apparatus described in the above (2) or (3), knowledge for judging an extracted opinion is automatically acquired and accumulated from a specific expression pattern in the answer sentence. A determination knowledge acquisition means 7-2 for performing the determination.

(7) In order to operate the computer as a natural sentence processing means for analyzing opinions provided by a plurality of opinion providers, the computerized data of the free description part of the opinion provider is provided from the provided opinions. Answer extraction means 2 for extracting the provided answer sentence, opinion extracting means 3 for extracting a word pair from the extracted answer sentence, opinion summing means 4 for summing up each extracted opinion content, and summarizing the result of each opinion content. A natural sentence processing program for functioning as the provided result providing means 5 is provided.

As a result, the following functions and effects can be obtained.

(1) Word pairs such as opinions, opinions, and impressions of the opinion provider can be automatically extracted as opinions from the answer sentence of the free description part submitted by a plurality of opinion providers. Opinions can be extracted accurately and in a short time from the answer sentence in the free description part.

(2) Searching a judgment knowledge database storing knowledge for judging whether the extracted opinion has given a positive evaluation or a negative evaluation regarding the object. By doing so, it is possible to accurately and automatically determine.

(3) When both positive and negative opinions are obtained from one answer sentence, the strength of the evaluation is compared and a comprehensive and detailed judgment result in the answer sentence is automatically obtained. be able to.

(4) Since the presentation information can be sequentially displayed in a plurality of graphs in which the whole set is changed, the analysis result can be presented in an easily understandable manner.

(5) Since the knowledge for judging the extracted opinion is automatically acquired from the answer result in the opinion questionnaire of a specific format, the opinion judgment can be made without preparing the knowledge for the judgment in advance. It can be carried out. Moreover, the more accurate the result, the more accurate the result can be obtained.

(6) Since the knowledge for determining the extracted opinion can be automatically acquired without using an opinion survey form of a specific format, the knowledge for determination in advance is prepared. Opinion judgment can be performed without doing. Moreover, the more accurate the result, the more accurate the result can be obtained.

(7) An answer extracting means for extracting, as a response sentence for providing an opinion, digitized data of a free description portion of the opinion provider from the provided opinion to the computer, and an opinion for extracting a word pair from the extracted answer sentence By installing a program that functions as an extracting unit, an opinion totalizing unit that totals each extracted opinion content, and a result providing unit that provides a totaling result for each opinion content, a normal computer can be installed from multiple opinion providers. A natural sentence processing function for analyzing the provided opinion can be performed.

[0030]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described with reference to FIGS. 1 is an embodiment of the present invention, FIG. 2 is an explanatory diagram of opinion extraction, FIG. 3 is an explanatory diagram of an opinion total result showing an example of an opinion total result in the present invention, and FIG. 4 is a total showing an example of a total result in the present invention. FIG. 5 is an explanatory diagram of a result, FIG. 5 is an explanatory diagram of a graph state of the present invention, and FIG. 6 is an analysis example of an answer sentence of the present invention.

In the figure, 1 is an input section, 2 is an answer extracting section, 3 is an opinion extracting section, 4 is an opinion totaling section, 5 is a result presenting section, 6 is an opinion judging section, and 7 is a judgment knowledge acquiring section. A first knowledge acquiring unit 7-1 and a second knowledge acquiring unit 7-2, 8 is a total knowledge database, 9 is a judgment knowledge database, 10 is a total result storage database, 20 is a processor, and 21 is a display. A is a questionnaire respondent and B is a questionnaire analyst.

The input section 1 is for the questionnaire respondent A to input each answer to a questionnaire sheet and to collect the answers in the apparatus. As a process for collecting this answer, for example, a paper questionnaire questionnaire can be directly distributed to the questionnaire target, or distributed by fax, etc.
This can be realized by distributing a survey form, entering a response, and collecting it using an electronic mail or a web bulletin board.

In the case of collecting a response to a questionnaire or collecting a response by facsimile, since the answer entry column is fixed, it can be detected by the coordinate position of the sheet. In the process of collecting such answers, the conventional means for collecting questionnaire results can be used as it is.

Note that the input unit 1 is not limited to the one that handles character information as described above, and it is also possible to collect the answer of the subject in the form of voice data by telephone response or the like. However, data on the Web using the Internet may be used, and any information transmission means can be used as a medium for answering. If translation means is provided, it can be input in a foreign language.

The answer extraction unit 2 converts the free description part in the collected individual questionnaire answer results into electronic text,
This is a process for extracting a response sentence. This processing is, for example, a method in which a questionnaire response on a collected paper medium is used by an operator on the side of the questionnaire to create digitized text using digitized data input means such as a keyboard, or an appropriate automatic digitizing means depending on the answer medium, For example, it is realized by a method of creating digitized text without using a character recognition device or a voice recognition device.

Depending on the means used in the questionnaire input unit 1, the collected answers may already be digitized. In such a case, the digitized free description part is directly used as the opinion extraction unit. 3 can be sent.

The opinion extracting unit 3 analyzes the digitized answer sentence created by the answer extracting unit 2 and extracts information on the answer sentence as an opinion.

The opinions extracted here include, for example, a product questionnaire on a notebook computer, which is included in the answer sentence "The body is small but the screen is large." Is large "is information that the questionnaire implementer wants to know at the time of questionnaire aggregation.

In this opinion extraction process, as shown in FIG. 2, the answer sentence is converted into a syntax tree by using the syntax analysis means 3-2 to extract its subject and predicate, and its binary relation "body-small". , "Screen-large" is set as the total processing unit.

The information to be extracted as opinions is
Not limited to the above-described "subject-predicate" pair, various things can be considered, such as extracting "screen-easy" from "easy-to-view screen" from the "modified-modifier" relationship.

The opinion is not limited to the above-described binary relation, but may be an opinion based on three or more relations in the syntax tree, for example, "subject-object-predicate". Furthermore, it is also possible to convert it into a semantic structure by the semantic interpreting means 3-1 and use it as an opinion.

For example, “♯screen → ♯beautiful” is represented. Here, ♯ is an operator representing the meaning of the word behind it, and represents not only the screen appearing in the text but also, for example, the entire synonym of the screen in the field, such as monitor, screen, screen, Shows the entire liquid crystal and the like. In addition, → is an operator representing an inference result, and in this example, indicates that “the screen is beautiful”. Note that all of these processes are performed using a computer, and can be realized without human intervention.

The opinion totalizing unit 4 performs a process of totalizing the frequency of appearance of each of the opinions extracted by the opinion extracting unit 3, for example, for each binary relation such as "body-small" or "screen-large". As a result, for example, as shown in FIG. 3, the frequency of appearance for each binary relation can be totaled.

As an aggregation method, for example, when opinions are expressed as the above-described binary relation, in addition to totalizing the number for each opinion, the first item (“body” if “body-small”) is used.
By performing tallying with only the second term or tallying with only the second term (“small” in the above example) and sending the tallying result to the result presenting unit 5, information presentation that matches the analysis purpose of the analyst B can be performed. It can be carried out.

Further, a tallying method having a function of tallying the results of absorbing the fluctuation of the expression of opinions can be used. This is a function of, for example, regarding the opinion extracted from the negative sentence, “body-not large”, summing up the opinion “body-not large” as the same opinion as “body-small”.

For example, a total knowledge database 8 including an antonym database showing an atonym of an antonym is connected to the opinion tallying unit 4, and if the second term of the opinion includes a negation, for example, an affirmative expression of "not large" This can be realized by searching this antonym database with "large", acquiring the antonym "small", combining it with the first term, and using it as "body-small" for counting.

In a similar manner, it is also possible to equip the total knowledge database 8 with a synonym database and to unify synonyms. In this process, for example, opinions such as “screen-beautiful” and “screen-beautiful” can be totaled as “screen-beautiful”.

In addition, a thesaurus, real world knowledge, and the like can be provided in the tallying knowledge database 8. Real-world knowledge means, for example, that the opinion "the previous one was good" is judged to be "the current one is bad", and the knowledge to be extracted from the answer sentence "the previous one was good for operability" The operability is bad. "

The result of performing the above processing is shown, for example, in a table shown in FIG. FIG. 3 relates to a notebook personal computer.
.. Are extracted, and the total number is 200, 50, 150..., “Body-small” and “screen-large” indicate favorable determination results, and “screen-large” “Small” indicates a negative determination result.

As the synonymous opinion of "body-small", "body-compact", "body-small" ...
However, as the synonymous opinion of "screen-large", "screen-wide" ..., but as the synonymous opinion of "screen-small", "screen-narrow", "screen-not large" ... It is shown that.

The data format sent to the result presentation unit 5 may be converted into a format handled by various table format software and sent in the form of a table as shown in FIG. 3, or may be sent in another format.
Note that all of the processing relating to these opinion counting units can be realized by a computer.

The result presenting unit 5 performs a process of presenting the totaling result created by the opinion totalizing unit 4 to the questionnaire analyst B. As a method of presenting the totaling result, the questionnaire analyst B may be presented to the questionnaire analyst B in the form of a table as shown in FIG. May be created and presented.

According to FIG. 4, it can be understood that the respondent is more interested in the target product that is larger or smaller than the case where the main body is clean or dirty.

As the presenting means, for example, it is necessary to be able to present a monitor chart to the questionnaire analyst on the display unit 21 shown in FIG. 1, but in the present invention, it is not always necessary to limit the chart information to the presenting means. Alternatively, the information may be converted into voice information using a speaker or the like as a result reading means and presented to the questionnaire analyst.

When presented as a graph, as shown in FIG. 5, the first chart is represented by a bar graph in which the length is the total number of opinions counted in the first term. For example, as shown in FIG. 4, the number of favorable opinions is 500 and the number of negative opinions is 20, as shown in FIG. 4, so a bar graph showing the total number of 520 is created, and a positive evaluation opinion (500) and a negative evaluation opinion (500) are prepared. 20) are presented in different colors, for example, when the questionnaire analyst designates the positive evaluation part with a pointing device such as a mouse, the percentages of the second items in the positive opinion are presented in a pie chart, and the circle When a pointing device is used to indicate an item in the graph, an answer sentence from which the opinion is extracted is displayed.

In FIG. 5, when the positive evaluation portion of the bar graph relating to the main body is pointed by the pointing device, a pie chart divided into three is displayed, and the area (a) is displayed.
, The answer sentence "small" is displayed. When the area (b) is pointed, "convenience" is displayed, and when the area (c) is pointed, "light" is displayed.

Further, when both a positive opinion and a negative opinion are obtained from one answer sentence, the strength of the evaluation is compared to estimate a comprehensive judgment result in the answer sentence. For this reason, when these two opinions come out, each is given a score in advance and weighted.

In this way, not only the total result of the opinion unit but also the total of the evaluation of the answer sentence can be totaled. Therefore, it is color-coded according to the ratio of the graph and first presented to the questionnaire analyst. By designating the portion of the ratio with a pointing device, it is possible to graphically present the ratio of opinions in an answer sentence that gives a comprehensive positive evaluation.

The opinion judging unit 6 judges whether each opinion information extracted by the opinion extracting unit 3 is a positive opinion or a negative opinion by using the opinion judgment knowledge stored in the judgment knowledge database 9. The processing is performed.

The opinion judgment knowledge stored in the judgment knowledge database 9 is, for example, a ternary relation such as “body-small-positive evaluation” when an opinion can be expressed by a binary relation as an example. This is information that can be described. In addition, in order to eliminate the object dependency in the judgment knowledge database 9 and to use the same database for questionnaires on other objects, information on the object is further added to the “body-small-positive evaluation, object: It may be described in terms of four terms such as "information equipment".

A judgment knowledge database storing such judgment knowledge is provided, and each extracted opinion, for example, “body-
Search this judgment knowledge database for "small"
For example, this processing can be realized by obtaining “body-small-positive evaluation”.

Further, for example, a score (weight) of +1 is given to a positive opinion and −1 is given to a negative opinion, and the sum of scores of a plurality of opinions obtained in one answer sentence is comprehensively calculated for one answer sentence. The evaluation of the object can be obtained for each answer sentence by regarding the evaluation as a proper evaluation.

Further, the judgment result is not binary, such as positive evaluation / negative evaluation, but multi-valued. For example, “strong affirmation” and “strong negation” are further added to the evaluation to obtain 4 values.
By giving a score according to the strength of the evaluation as a value, a more accurate determination can be achieved. It is also possible to express each evaluation value by an evaluation function and calculate a comprehensive evaluation by performing synthesis based on bundle value logic, so-called “fuzzy inference”.

The judgment knowledge acquiring section 7 automatically acquires judgment knowledge without using any manual operation. The judgment knowledge used in the opinion judgment unit 6 is object-dependent knowledge, and maintenance of the judgment knowledge is costly. In order to solve this problem, it is only necessary to automatically acquire the judgment knowledge from the questionnaire answer sentence.

The first method is to automatically acquire and accumulate knowledge for judging an extracted opinion automatically from a questionnaire response result in a questionnaire questionnaire in a judgment format.

In the initial stage of operation, since sufficient judgment knowledge is not stored in the judgment knowledge database 9, each question of the questionnaire is, for example, “for this product,
Please fill in the points that you think are good. ", And extract the opinions from the answer sentence and combine the expected evaluation with the answer. Is stored in the judgment knowledge database as judgment knowledge.

If the attributes to be evaluated for the object are known in advance, the question is “Select the following items of this product that you felt from the following, and explain the reason why you felt so. Please fill in. "
Attributes to be evaluated such as "keyboard" are arranged, and the answer part is provided with two items: an option entry item "very good-good-normal-bad-very bad" and a free description item to enter the reason. From the results of the electronic questionnaire results, from the items such as "screen" and terms from the free description part, for example, the description "because it is large and easy to see"
“Large” and “easy to see” are extracted, an opinion is created based on the combination, and the evaluation value of the selected portion, for example, “good” is combined to create judgment knowledge. For example, "screen-large-
"Positive evaluation" and "screen-easy-to-view-positive evaluation". By storing the knowledge acquired by such processing in the judgment knowledge database 9, it becomes possible to use it in the subsequent processing. The above processes are performed by the first determination knowledge acquisition unit 7-1 shown in FIG.

As a second method, as described above, judgment knowledge is acquired without setting a special question at the beginning of operation, and the second judgment knowledge acquisition unit 7- shown in FIG. Reference numeral 2 denotes a processing unit.

This is, for example, “the screen can be easily viewed”.
From the answer sentence, two opinions, "screen-easy to see" and "screen-good" can be extracted, but if one of the second terms is a word representing a previously given evaluation, in this example, " Since “good” indicates an evaluation given in advance, this knowledge is expressed by a binary relation of “good-positive evaluation”. Then, by merging the two, it is possible to obtain the determination knowledge information “screen-easy-to-view-positive evaluation”.

This method not only extracts judgment knowledge, but also
It is also possible to acquire words representing evaluations that have not been given to the device in the initial state of operation. It is assumed that, for example, the knowledge “good-positive evaluation” is given to the apparatus in advance, and the determination knowledge “screen-easy-to-view-positive evaluation” is acquired from the above example sentence.

Next, two opinions such as “screen-easy to see” and “screen-great” are extracted from the answer sentence “screen is easy to see and wonderful”, but “screen-easy to understand” is affirmed from the acquired judgment knowledge. Since the evaluation is a positive evaluation, it is possible to obtain new knowledge such as "excellent-positive evaluation". As described above, the fact that “easy to see” and “great” is a positive evaluation is stored in the determination knowledge database 9 and is used for subsequent processing, thereby realizing a knowledge learning function.

All of the above-described processing can be performed by a computer without human intervention. Note that the first method,
The determination knowledge acquisition method of the second method is not exhaustive, but may be a combination of both.

The tallying knowledge database 8 stores data necessary for tallying answer sentences. For example, an antonym database, a synonym database, a thesaurus, real world knowledge, and the like are stored.

The judgment knowledge database 9 stores data necessary for judging whether the answer sentence is a positive (favorable) opinion or a negative opinion. For example, the part-of-speech, opinion judgment knowledge, " 1st term + 2nd term + evaluation ”,“ 1st term
Item + second item + evaluation + target field ”, points (weight) for each opinion when a positive opinion and a negative opinion coexist are stored.

The counting result storage database 10 temporarily stores the counting results obtained by the present invention, and stores the counting data shown in FIGS. 3 and 4, the display data shown in FIG. 5, and the like. is there.

The processor 20 constitutes the natural processing apparatus of the present invention, and includes, for example, an answer extracting unit 2, an opinion extracting unit 3, an opinion totaling unit 4, a result presenting unit 5, an opinion judging unit 6, and a judgment knowledge acquiring unit. 7 and the like.

The display section 21 is for the questionnaire analyst B to display the processing result of the present invention and to see the processing result, and is composed of a display, a printer, and the like.

The operation of the embodiment of the present invention shown in FIG. 1 will be described with reference to the flowchart shown in FIG.

S1. The answer to the questionnaire from the questionnaire respondent A is input to the input unit 1 by facsimile or electronic information, for example. The input unit 1 sends the answer result to the answer extracting unit 2.

S2. The answer extraction unit 2 performs a process of converting the answer result, particularly the free description part, into an electronic text and extracting it as an answer sentence. Then, the computerized answer sentence is sent to the opinion extracting unit 3.

S3. The opinion extracting unit 3 morphologically analyzes the digitized answer sentence, converts it into a parse tree using the parsing means 3-2, and creates a binary or ternary relation,
Extract opinions from the answer sentence. Then, this is sent to the opinion counting unit 4.

S4. The opinion tallying unit 4 is provided with the opinion extracting unit
Are collected, for example, for each binary relation, the appearance frequency of the extracted opinions. At this time, the opinion judgment unit 6 judges whether the opinion extracted by the opinion extraction unit 3 is a positive opinion or a negative opinion, using the opinion judgment knowledge stored in the judgment knowledge database.

S5. The opinion counting unit 4 counts the appearance frequencies of the opinions extracted by the opinion extraction unit 3 in this manner, and determines whether the opinion extracted by the opinion extraction unit 3 via the opinion determination unit 6 is positive or negative. Further, the synonym database is accessed and the synonymous opinion is recognized. Then, as shown in FIG. 3, an opinion totaling table is created.

S6. When the opinion tally created in this way is sent to the result presenting unit 5, the result presenting unit 5 processes this and presents the questionnaire analyst B in an easy-to-analyze form. For example, as shown in FIG. 4, the data may be presented in the form of a table, or a table or a graph matching the analysis purpose of the questionnaire analyst may be created and presented. Alternatively, as shown in FIG. 5, when specified by a pointing device, the portion may be presented so as to be displayed in more detail.

When the answer sentence shown in FIG. 6 is input, in the present invention, the opinion extractor 3 and the opinion determiner 6 morphologically analyze the answer sentence, and the subject-predicate of the first half sentence is "baseball". It recognizes that "ga" is a negative conjunction and recognizes "baseball-thought-negation".

The subject-predicate of the latter sentence is "golf-
"Good" and a positive evaluation of "good" recognizes "gokufu-good-positive". In this way, the answer sentence of FIG. 6 obtains "not baseball-golf".

In the above description, the present invention has been described with respect to an example of processing of a questionnaire response sentence, but the present invention is of course not limited to this.

The answer extracting unit 2, the opinion extracting unit 3, the opinion totaling unit 4, the result presenting unit 5, the opinion judging unit 6, the judgment knowledge acquiring unit 7 and the like can be constituted by a program, for example.

The embodiments of the present invention are described in Appendix.

(Supplementary Note 1) In a natural sentence processing apparatus that analyzes opinions provided by a plurality of opinion providers, digitized data of a free description portion of the opinion provider is extracted from the provided opinions as an answer providing answer. Answer extracting means, an opinion extracting means for extracting a word pair from the extracted answer sentence, an opinion totalizing means for totalizing each extracted opinion content, and a result providing means for providing a totaling result for each opinion content. A natural sentence processing device, comprising:

(Supplementary Note 2) In the natural sentence processing apparatus according to Supplementary Note 1, an opinion judging means for judging whether the extracted opinion gives a positive evaluation or a negative evaluation, and an opinion judgment A natural sentence processing apparatus comprising a judgment knowledge database for storing knowledge for performing

(Supplementary Note 3) In the natural sentence processing apparatus according to Supplementary Note 2, when two positive opinions and negative opinions are obtained in one answer sentence, the strength is determined, and the overall answer sentence is determined. A natural sentence processing device comprising opinion judgment means for judging positive or negative.

(Supplementary Note 4) The natural sentence processing apparatus according to Supplementary Note 1 or 2, further comprising presentation means for displaying presentation information as a plurality of graphs in which a whole set is changed. .

(Supplementary Note 5) In the natural sentence processing apparatus according to Supplementary Notes 2 and 3, the knowledge for judging the extracted opinion is automatically acquired and accumulated from the answer result in the opinion survey form of a specific format. A natural sentence processing apparatus characterized by comprising judgment knowledge acquiring means for performing judgment.

(Supplementary Note 6) In the natural sentence processing apparatus according to Supplementary Notes 2 and 3, especially, knowledge for judging an extracted opinion is automatically acquired from a specific expression pattern in an answer sentence and accumulated. A natural sentence processing device comprising a judgment knowledge acquiring means.

(Supplementary Note 7) In order for the computer to operate as a natural sentence processing means for analyzing opinions provided by a plurality of opinion providers, digitized data of a free description portion of the opinion provider is provided from the provided opinions. Provide answer extraction means for extracting the answer sentence provided by the opinion, opinion extraction means for extracting a word pair from the extracted answer sentence, opinion counting means for counting each extracted opinion content, and provide a tally result for each opinion content A natural sentence processing program to function as a result providing means.

[0097]

According to the present invention, the following effects can be obtained.

(1) Word pairs such as opinions, views, and impressions of the opinion provider can be automatically extracted as opinions from the answer sentence of the free description part submitted by a plurality of opinion providers, so that a large number of freedoms can be obtained. Opinions can be extracted accurately and in a short time from the answer sentence in the description part.

(2) Searching a judging database storing knowledge for judging whether the extracted opinion gives a positive evaluation or a negative evaluation regarding the object or not. By doing so,
Automatic determination can be made.

(3) When both positive and negative opinions are obtained from one answer sentence, the evaluation strengths are compared to automatically obtain comprehensive and detailed judgment results in the answer sentence. be able to.

(4) Answer extraction means for extracting, from the provided opinion, digitized data of the free description portion of the opinion provider as an answer sentence provided by the opinion provider, and an opinion for extracting a word pair from the extracted answer sentence By installing a program that functions as an extracting unit, an opinion totalizing unit that totals each extracted opinion content, and a result providing unit that provides a totaling result for each opinion content, a plurality of opinion providers can be installed on a normal computer. A natural sentence processing function of analyzing the provided opinion can be performed.

[Brief description of the drawings]

FIG. 1 is an embodiment of the present invention.

FIG. 2 is an explanatory diagram of opinion extraction according to the present invention.

FIG. 3 is an explanatory view of an opinion totaling result in the present invention.

FIG. 4 is an explanatory diagram of a counting result in the present invention.

FIG. 5 is an explanatory diagram of a graph state in the present invention.

FIG. 6 is an example of answer sentence analysis in the present invention.

FIG. 7 is an explanatory diagram of an operation state of the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Input part 2 Answer extraction part 3 Opinion extraction part 4 Opinion totalization part 5 Result presentation part 6 Opinion judgment part 7 Judgment knowledge acquisition part 8 Total knowledge database 9 Judgment knowledge database 10 Total result storage database 20 Processor 21 Display part

Claims (4)

[Claims] Claims: 1. A natural sentence processing apparatus for analyzing opinions provided by a plurality of opinion providers, wherein a response extracting electronic data of a free description portion of the opinion provider from the provided opinions as an answer providing answer statement. Extraction means, opinion extraction means for extracting word pairs from the extracted answer sentences, opinion totalization means for totalizing each extracted opinion content, and result providing means for providing a totaling result for each opinion content A natural sentence processing device characterized by the above-mentioned. 2. The natural sentence processing apparatus according to claim 1, wherein opinion judgment means for judging whether the extracted opinion gives a positive evaluation or a negative evaluation, and judges the opinion A natural sentence processing device comprising a judgment knowledge database for storing knowledge for the purpose. 3. In the natural sentence processing apparatus according to claim 2, when two positive opinions and negative opinions are obtained in one answer sentence, the strength is determined, and the overall answer sentence is determined. A natural sentence processing device comprising an opinion judging means for judging affirmation / negation. In order to operate a computer as a natural sentence processing means for analyzing opinions provided by a plurality of opinion providers, electronic opinion data of a free description portion of the opinion provider is provided from the provided opinions. Extraction means to extract word pairs from the extracted answer sentences, Opinion extraction means to extract word pairs from the extracted answer sentences, Opinion summarizing means to sum up each extracted opinion content, Provision of results that provide the totaling result for each opinion content Natural sentence processing program to function as a means.