JP4293145B2 - Word-of-mouth information determination method, apparatus, and program - Google Patents

Description

  The present invention relates to a word-of-mouth information determination method, apparatus, and program, and in particular, from a situation where all types of information such as the Internet are mixed, the sentence included in the information is analyzed, and an experience related to a certain thing The present invention relates to a word-of-mouth information determination method, an apparatus, and a program using experience expressions, which can distinguish and determine “user's voice, that is, word-of-mouth information” in which talks, comments used, and complaints are written.

  In recent years, many people have published their opinions on the Internet regarding various matters such as hotels, travel, and electrical appliances. This public place includes a bulletin board where opinions of various people gathered, such as a shopping site itself, a weblog where personal opinions are gathered, and a personal homepage. Such user's voice is highly useful as support for marketing, product development, product purchase examination, etc. as user's voice and reputation (hereinafter referred to as word-of-mouth information).

  Currently, the following technologies have been proposed for analyzing word-of-mouth information, in which specific products and information collection destinations (URLs, etc.) are limited, and descriptive expressions depending on the subject matter are extracted from sentences. .

  As a first technique, there is a technique for extracting an opinion from a Web page on the Internet by using a method for calculating a proximity between a product name input by a user and an evaluation expression prepared in advance as a dictionary (see, for example, Non-Patent Document 1). . The evaluation expression used here is an expression indicating a person's evaluation of things, and the dictionary is prepared for each product category created in advance. Examples of the product category include “book”, “computer”, and “car”.

  On the other hand, as a second technique, in the Internet search, various web pages such as an official page of a company related to a certain keyword, a shopping site, and word-of-mouth information are often output without distinction. In order to navigate from such a vast amount of information to information desired by the user, a method of analyzing the output result has been proposed. The method analyzes a web page document collected by a search engine, extracts a characteristic specific expression such as a person name and a place name from the document, and presents it to a user as an index for a search result. Is a technique for classifying (see, for example, Non-Patent Document 2).

  As a third technique, there is a proposal for classifying the content type based on the URL of the search result. The Web page B including the link (URL) of the Web page A output as the search result is analyzed. This is a technique for analyzing the surrounding information (anchor text) in which the link of A is embedded in B and estimating the format of the original Web content. The user selects a function of the content at the time of information search, such as the function of the content: “satisfy the purpose by obtaining the web content itself”, “satisfy the purpose by guide information such as a book or seminar indicated by the web content”. There is a need to. The types of content here are “paper”, “commentary article”, “news”, “bulletin board”, “diary”, “site”, “link collection”, and “guide” (for example, see Non-Patent Document 3).

Moreover, when using said 1st technique and 3rd technique, it is necessary to perform the morphological analysis of a sentence. The morpheme analysis technology is a Japanese morpheme analysis system that reads homomorphic words using word co-occurrence. Since co-occurrence of words is easy to identify for a person, it is difficult to make a mistake and the accuracy of the system can be surely improved (see, for example, Non-Patent Document 4).
Kenji Tateishi et al., "Reputation Information Retrieval from the Internet", IPSJ Research Report, 2001-NL-144, pp.75-82, 2001 Hiroyuki Toda et al., “Proposal of Label Oriented Navigation Method Using Characteristic Specific Expression”, Information Processing Society of Japan, Vol.2004-no.45, 2004-FI-75, pp.99-106 Noriaki Kawamae et al., “Proposal on Structuring Search Results Focusing on Web Content Functions”, Institute of Population Intelligence SGI-SWO-A303-10 Takeshi Tsuji et al., "Japanese Morphological Analysis System Considering Conservativeness", Report of the Natural Language Processing Study Group of Information Processing Society of Japan, 117-09, vol.1997-no.004, pp.59-66

  In this specification, “review information” is defined as follows. Word-of-mouth information is information about a certain subject (for example, a digital video camera, a hotel, a restaurant, a TV drama, an actor, an English conversation class, a sightseeing spot, a case, etc.). A person who has experienced or experienced describes their opinions such as their experiences and their experiences. For example, from the sentence, “I recently went to a new restaurant in Shibuya. The restaurant was very delicious with rumored carbonara.” From the sentence “I tried it” and “It was delicious.” It can be said that it is word-of-mouth information for XX restaurant. This kind of word-of-mouth information can be used as a reference for product development and product improvement in companies, and when general users see it, it is useful information useful for consideration when selecting purchases, travel destinations, etc. .

  When determining whether or not a certain sentence is word-of-mouth information, the above-described conventional first to third methods have the following problems.

  In the first technology, the following two types are treated as reputation information without distinction.

1) Opinions and evaluations predicted and judged by the writer under the conditions presented by other people's experiences, reports, machine specifications, etc .:
(Example) “Product A is cheap for specs”, “In B, the number of pick-ups has increased recently and the security is poor”
2) Opinions and evaluations of people who actually experienced the subject matter (synonymous with the definition of word-of-mouth information in the present invention)
(Example) “When I used Product A, the display was dark”, “When I went to B, the city was not so safe”
However, in actual usage scenes such as purchasing products, considering travel destinations, and selecting dieting methods, opinions based on actual experiences are often considered more important. In the first technique, it is necessary to create a specific product category and a dictionary corresponding to the category beforehand. At present, opinions on various things are published on the Internet, and it is very expensive to create all the dictionaries according to it.

  Further, in the second technique, word-of-mouth information cannot be determined because classification is performed using a unique expression such as a person name or a company name.

  In the third technique, the bulletin board and diary types can be classified, but classification is not performed from the viewpoint of “word-of-mouth information” partially included therein.

  Further, even if the information satisfies the above-described definition of the review information, if there is only one opinion, there is a problem that the bias of the opinion of the individual occurs in the review information. However, it is difficult for the viewer to determine whether or not the author has a preference, and thus it is difficult to determine whether or not it is reference information for the viewer.

  The present invention has been made in view of the above points, and does not limit the fields and objects, and analyzes sentences contained in the information from situations where all kinds of information such as the Internet are mixed. Reviews that can be identified by distinguishing "user's voice, that is, word-of-mouth information" that contains experiences, impressions, complaints, etc. related to certain things by using the presence or absence of experience expressions and evaluation expressions It is an object to provide an information determination method, apparatus, and program.

  FIG. 1 is a principle configuration diagram of the present invention.

The present invention (Claim 1) is a word-of-mouth information determination device for determining word-of-mouth information including experiences and complaints about a certain thing from an input text, and complaints,
An evaluation expression storage means 20 for storing an evaluation expression that is a phrase that evaluates the quality of things;
Experience expression storage means 21 for storing experience expressions, which are phrases representing the use or experience / experience of things;
The input text is referred to the evaluation expression storage means 20 to determine whether or not an evaluation expression that is a phrase that evaluates the quality of things is described, and the determination result is stored in the data storage means 501. Evaluation expression checking means 520 for performing,
The input text is referred to the experience expression storage means 21 to determine whether or not an experience expression that is a phrase that represents the use or experience / experience of a thing is described, and the determination result is stored in the data storage means 501. Experiential expression inspection means 530 stored in
When the information stored in the data storage unit 501 is read out, the evaluation expression checking unit 520 determines that the evaluation expression is described, and the experience expression checking unit 530 determines that the experience expression is described. And word-of-mouth determination means 540 for determining whether or not the word-of-mouth is a word-of-mouth by a word-of-mouth determination function represented by an evaluation expression and a calculation formula based on the experience expression.

The present invention (Claim 2) is a word-of-mouth information determination apparatus that determines word-of-mouth information including experiences and complaints about a certain thing, complaints, and complaints from the input text or the structure of a Web page,
An evaluation expression storage means for storing an evaluation expression, which is a phrase that evaluates the quality of things,
Experience expression storage means for storing experience expressions, which are phrases representing the use or experience / experience of things,
A set of one or more sentences and a name or symbol including the writer's name or nickname of the body, date information describing the body, in the input text or web page structure, or , A multi-writer inspection means for inspecting whether the text, the name or the set of symbols appears repeatedly;
An evaluation expression that refers to the evaluation expression storage means, determines whether or not an evaluation expression that is a word that evaluates the quality of things is described in the text, and stores the determination result in the data storage means Inspection means;
An experience that refers to experience expression storage means, determines whether the experience expression, which is a phrase that expresses the use or experience / experience of things, is described in the text, and stores the determination result in the data storage means Expression inspection means;
When the determination result is read from the data storage means, it is determined that the evaluation expression is described by the evaluation expression inspection means, and the experience expression is determined by the experience expression inspection means, the evaluation expression And a word-of-mouth determination means for determining whether or not the word-of-mouth is a word-of-mouth by a word-of-mouth determination function represented by a calculation formula based on the experience expression.

The present invention (Claim 3) is a word-of-mouth information determination apparatus for determining word-of-mouth information including experiences and complaints about a certain thing from the input text, complaints,
An evaluation expression storage means for storing an evaluation expression, which is a phrase that evaluates the quality of things,
Experience expression storage means for storing experience expressions, which are phrases representing the use or experience / experience of things,
An evaluation that refers to the evaluation expression storage means, determines whether or not an evaluation expression that is a phrase that evaluates the quality of things is described in the input text, and stores the determination result in the data storage means Expression inspection means;
Referring to the experience expression storage means, determine whether the input text contains an experience expression that is a phrase that expresses the use, experience, or experience of things, and stores the determination result in the data storage means An experiential expression inspection means,
When the determination result is read from the data storage means, it is determined that the evaluation expression is described by the evaluation expression inspection means, and the experience expression is determined by the experience expression inspection means, the evaluation expression And a word-of-mouth rate calculating means for calculating a word-of-mouth rate, which is a ratio occupied by word-of-mouth information in the entire input text, by a word-of-mouth determination function represented by a calculation formula based on the experience expression.

The present invention (Claim 4) is a word-of-mouth information determination apparatus that determines word-of-mouth information including experiences and complaints about a certain thing in the input text or the structure of the Web page, complaints,
An evaluation expression storage means for storing an evaluation expression, which is a phrase that evaluates the quality of things,
Experience expression storage means for storing experience expressions, which are phrases representing the use or experience / experience of things,
A set of one or more sentences and a name or symbol including the writer's name or nickname of the body, date information describing the body, in the input text or web page structure, or , A multi-writer inspection means for inspecting whether the text, the name or the set of symbols appears repeatedly;
An evaluation expression that refers to the evaluation expression storage means, determines whether or not an evaluation expression that is a word that evaluates the quality of things is described in the text, and stores the determination result in the data storage means Inspection means;
Experience expression that refers to experience expression storage means, determines whether experience expression that is a phrase that expresses the use or experience / experience of things in the text is described, and stores the determination result in data storage means Inspection means;
When the determination result is read from the data storage means, it is determined that the evaluation expression is described by the evaluation expression inspection means, and the experience expression is determined by the experience expression inspection means, the evaluation expression And a word-of-mouth rate calculation means for calculating a word-of-mouth rate, which is a ratio occupied by word-of-mouth information, of the input text or the entire Web page by a word-of-mouth determination function represented by a calculation formula based on the experience expression.

  FIG. 2 is a diagram for explaining the principle of the present invention.

The present invention (Claim 5) is a review information determination method for determining review information including experiences, complaints, and complaints related to certain things from the input text.
Referring to an evaluation expression storage means for storing an evaluation expression that is a phrase that evaluates the quality of things, it is determined whether or not the evaluation expression is described in the input text, and the determination result is stored in data An evaluation expression checking step (step 1) stored in the means;
Refers to experience expression storage means for storing experience expressions, which are phrases representing the use or experience / experience of things, determines whether the experience expression is described in the input text, and the determination result as data An experiential expression inspection step (step 2) to be stored in the storage means;
When the determination result is read from the data storage means, it is determined that the evaluation expression is described in the evaluation expression inspection step, and the evaluation expression is determined to be described in the experience expression inspection step, the evaluation expression And a word-of-mouth determination step (step 3) for determining whether or not the word-of-mouth is a word-of-mouth by a word-of-mouth determination function represented by a calculation formula based on the experience expression.

The present invention (Claim 6) is a word-of-mouth information determination method for determining word-of-mouth information including experiences, complaints, and complaints related to certain things from the input text or the structure of a Web page.
A set of one or more sentences and a name or symbol including the writer's name or nickname of the body, date information describing the body, in the input text or web page structure, or A multi-author inspection step for inspecting whether the text, the name or symbol pair repeatedly appears;
Refers to the evaluation expression storage means for storing the evaluation expression, which is a phrase that evaluates the quality of things, determines whether the evaluation expression is described in the text, and stores the determination result in the data An evaluation expression checking step stored in the means;
Refers to the experience expression storage means for storing the experience expression that is a phrase representing the use or experience / experience of things, determines whether the experience expression is described in the text, and the determination result as a data storage means And the experience expression inspection step to be stored in
When the determination result is read from the data storage means, it is determined that the evaluation expression is described in the evaluation expression inspection step, and the evaluation expression is determined to be described in the experience expression inspection step, the evaluation expression And a word-of-mouth determination step for determining whether or not the word-of-mouth is a word of mouth by a word-of-mouth determination function represented by a calculation formula based on the experience expression.

The present invention (Claim 7) is a review information determination method for determining review information including experiences, complaints, and complaints related to certain things from the input text.
Referring to an evaluation expression storage means for storing an evaluation expression that is a phrase that evaluates the quality of things, it is determined whether or not the evaluation expression is described in the input text, and the determination result is stored in data An evaluation expression checking step stored in the means;
Refers to experience expression storage means for storing experience expressions, which are phrases representing the use or experience / experience of things, determines whether the experience expression is described in the input text, and the determination result as data An experiential expression inspection step stored in a storage means;
When the determination result is read from the data storage means, it is determined that the evaluation expression is described in the evaluation expression inspection step, and the evaluation expression is determined to be described in the experience expression inspection step, the evaluation expression And a word-of-mouth rate calculation step of calculating a word-of-mouth rate, which is a ratio occupied by word-of-mouth information in the entire input text, using a word-of-mouth determination function represented by a calculation formula based on the experience expression.

The present invention (Claim 8) is a word-of-mouth information determination method for determining word-of-mouth information including experience stories, impressions used, and complaints related to certain things from the input text or the structure of a Web page.
A set of one or more sentences and a name or symbol including the writer's name or nickname of the body, date information describing the body, in the input text or web page structure, or A multi-author inspection step for inspecting whether the text, the name or symbol pair repeatedly appears;
Refers to an evaluation expression storage means for storing an evaluation expression that is a word that evaluates the quality of things, determines whether or not the evaluation expression is described in the text, and determines the determination result as a data storage means An evaluation expression checking step to be stored in
Refers to the experience expression storage means for storing the experience expression that is a phrase representing the use or experience / experience of things, determines whether the experience expression is described in the text, and the determination result as a data storage means And the experience expression inspection step to be stored in
When the determination result is read from the data storage means, it is determined that the evaluation expression is described in the evaluation expression inspection step, and the evaluation expression is determined to be described in the experience expression inspection step, the evaluation expression And a word-of-mouth rate calculation step of calculating a word-of-mouth rate, which is a ratio occupied by the word-of-mouth information, of the input text or the entire Web page by a word-of-mouth determination function represented by a calculation formula based on the experience expression.

The present invention (Claim 9) is a word-of-mouth information determination program for determining word-of-mouth information including input from an input text or an experience story about a certain thing from the structure of a Web page, a complaint,
An evaluation expression storage means for storing an evaluation expression, which is a phrase that evaluates the quality of things,
9. A process for realizing the word-of-mouth information determination method according to claim 5 on a computer comprising experience expression storage means for storing an experience expression that is a phrase representing use or experience / experience of things. It is a program to let you.

The present invention (Claim 10) stores a word-of-mouth information determination program for determining word-of-mouth information including experience stories, impressions used, and complaints related to certain things from the input text or the structure of a Web page. A medium,
An evaluation expression storage means for storing an evaluation expression, which is a phrase that evaluates the quality of things,
9. A process for realizing the word-of-mouth information determination method according to claim 5 on a computer comprising experience expression storage means for storing an experience expression that is a phrase representing use or experience / experience of things. A storage medium storing a program to be executed.

  Word-of-mouth information is information about a certain subject (for example, a digital video camera, a hotel, a restaurant, a TV drama, an actor, an English conversation class, a sightseeing spot, a case, etc.). A person who has experienced or experienced describes their opinions such as their experiences and their experiences. For example, from the sentence, “I recently went to a new restaurant in Shibuya. The restaurant was very delicious with rumored carbonara.” From the sentence “I tried it” and “It was delicious.” It can be said that it is word-of-mouth information for XX restaurant. Thus, it can be said that the word-of-mouth information includes not only the subject matter but also the experience expression as described above (eg: I tried it) and the evaluation expression (eg: it was delicious) that evaluates it. . In addition, since the word-of-mouth information is described with respect to all things, it is desirable that it can be determined without depending on the field or the object.

  According to the present invention, it is possible to distinguish two kinds of reputation information mentioned in the column of the problem to be solved by the above-mentioned invention by using the presence / absence of the experience expression that has not been used so far. It is possible to determine whether the information is “information in which opinions based on the user's experience / experience are written”. This expression of experience does not depend on the subject matter of the word of mouth information. By confirming whether both the evaluation expression and the experience expression are described, it is possible to determine whether or not the information is word-of-mouth information regardless of what the object is.

  In addition to the above, it is also possible to determine, as word-of-mouth information, a Web page or text in which word-of-mouth information of a plurality of people is gathered for a certain target object. Normally, when reviewing word-of-mouth information, it is rarely satisfied to see only one word-of-mouth information of a certain individual, and a plurality of word-of-mouth information is referred to to obtain multifaceted information. In other words, due to the effects described above, the viewer can see the opinions of a plurality of people at a time, and the trouble of referring to a plurality of word-of-mouth information pages can be saved. In this determination method, there are two pieces of information, ie, “name” indicating a name or nickname for identifying the author and “text” composed of one or more sentences, or in addition to this information “ Three pieces of information “date information” are determined as information for one person. By adding date information and name, the reliability of the information can be improved and the newness of the information can be known. In addition, this can also avoid the above-described problem in which the bias of one person's opinion occurs in the word-of-mouth information.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  In the following, “word-of-mouth information” is information about a certain subject (for example, a digital video camera, a hotel, a restaurant, a TV drama, an actor, an English conversation class, a sightseeing spot, a case, etc.) It is defined that a person who has experienced / experienced something in some way describes an opinion such as his / her experiences and impressions, and when there is input such as text or web page, whether or not it is word-of-mouth information The determination device will be described as a “word-of-mouth determination device”.

  FIG. 3 is a diagram for explaining the outline in one embodiment of the present invention.

When text or a web page is input to the review determination device,
1) A “writer's name” and B “text consisting of one or more sentences” and C “date” are one review set, and this A, B, C set or A, B Determine whether the set is repeated.

  2) Extract a group of sentences determined to be the text from the word-of-mouth set determined in 1).

  3) Check whether the evaluation expression is described in the body (text) extracted in 2).

  4) Check whether experience expression is described in the text.

  5) From the results of 3) and 4) above, it is determined whether or not it is word-of-mouth information, and the determination result is output.

  FIG. 4 shows the structure of the word-of-mouth information determination apparatus in one embodiment of the present invention.

  The word-of-mouth determination device shown in FIG. 1 includes a computer 10, an evaluation expression storage unit 20 and an experience expression storage unit 21 connected to the computer 10 via a network 40.

  The computer 10 includes a data storage unit 501 including a memory such as a RAM, a ROM, and a magnetic disk, a CPU, a display unit 11 using a display, and an instruction input unit 12 including a mouse and a keyboard. Has a multi-writer inspection processing unit 500, a body text extraction processing unit 510, an evaluation expression inspection processing unit 520, an experience expression inspection processing unit 530, and a word-of-mouth information determination processing unit 540, which are realized by a software program executed by.

  In the data storage unit 501, data is stored and read out by the multi-writer inspection processing unit 500, the text extraction processing unit 510, the evaluation expression inspection processing unit 520, the experience expression inspection processing unit 530, and the word-of-mouth information determination processing unit 540.

  The evaluation expression storage unit 20 stores words and phrases that express a good image or a bad image with respect to things. An example of a good image is “good, beautiful, delicious, interesting, fun, comfortable, convenient, gorgeous, abundant, likes”. Examples of bad images include “bad, dirty, bad, uninteresting, boring, inconvenient, noisy, annoying, tricky, worst, difficult” and the like. These apply mainly to adjectives, adjective verbs and nouns.

  The experience expression storage unit 21 stores words and phrases that represent experience and experience of things. For example, “I went, tried, executed, felt, realized, challenged”. These apply mainly to the past forms of verbs “~ ta”, “~”, “~”, “has been”, and so on. In the experiential expression storage unit 21, words or phrases corresponding to experiential expressions {actually realized, challenged, etc.} may be stored, or expression forms {already tried, etc.} or morphological analysis may be performed. The later part of speech combination rules {verb stem + verb inflection ending + verb suffix + verb stem continuation “look” + verb suffix ending “is” etc.}, or mixed information is stored There may be. An example of experience expression is shown in FIG.

  FIG. 6 is a flowchart of the outline operation of the word-of-mouth information determination process according to the embodiment of the present invention.

Step 100) Multiple writer inspection process:
When a text or a web page is input to the multi-author inspection processing unit 500, a text composed of one or more sentences and an author name or nickname of the text are shown in the text or web page structure. It checks whether or not three pieces of information of name or symbol and date information describing the body, or two pieces of information of body and name or symbol appear repeatedly, and stores the result in the data storage unit 501.

  FIG. 7 shows an example of word-of-mouth information input in one embodiment of the present invention. FIG. 4A is an example of a Web page (html file), and FIG. 4B is an example of text. As shown in FIG. 8, the relationship between the input document and the text, text, and text is one set of text composed of the author name, date, and one or more texts.

Step 200) Text extraction processing:
In the text extraction processing unit 510, the information in which the three pieces of information in step 100 repeatedly appear is read from the data storage unit 501, the text is extracted from the information, and stored in the data storage unit 501.

Step 300) Evaluation expression inspection processing:
The evaluation expression inspection processing unit 520 reads out the text extracted in step 200 from the data storage unit 501, checks whether or not the evaluation expression is described therein by referring to the evaluation expression storage unit 20, and the result Is stored in the data storage unit 501.

Step 400) Experience expression inspection processing:
The experience expression inspection processing unit 530 reads out the text extracted in step 200 from the data storage unit 501 and checks whether the experience expression is described therein by referring to the experience expression storage unit 21, and the result Is stored in the data storage unit 501.

Step 500) Review determination processing:
The word-of-mouth information determination processing unit 540 reads the test results in step 300 and step 400 from the data storage unit 501 and describes the evaluation expression and the experience expression, and determines whether or not it is a word-of-mouth by a predetermined word-of-mouth determination function. Alternatively, the review rate is determined.

[First Embodiment]
In the present embodiment, an example is shown in which it is determined whether or not the input document is word-of-mouth information and whether or not there is a word-of-mouth expression in the sth body. In this embodiment, presence / absence of word-of-mouth expression is output.

  First, the multi-writer inspection process (step 100) in the first embodiment will be described.

  FIG. 9 shows an example of a word-of-mouth information input html structure according to the first embodiment of the present invention.

  As shown in FIG. 3, the multiple writer inspection processing unit 500 takes A “writer's name”, B “text composed of one or more sentences”, and C “date” as one review set, and the repetition is repeated. Determine if there is. However, C “date” is not always necessary, and it is possible to form a word-of-mouth set with A and B. Here, the order of appearance of A, B, C or A, B is not limited. Below, the example in case three, A, B, and C exist is demonstrated. The order of appearance may be any of A → B → C, B → C → A, C → A → B, A → C → B, B → A → C, C → B → A. However, only one form appears repeatedly. For example, in the case of repetition of A → B → C, if C → B → A appears in the middle, it means that a different format has appeared, and it is considered that the repetition has ended. There are various determinations of repetition, but here, it starts from searching for A or C. 7 and 8 are examples of A → C → B.

  FIG. 10 is a flowchart of the multi-writer inspection process according to the first embodiment of the present invention, and shows an example starting from a name.

  Step 101) The multi-author inspection processing unit 500 searches for a character string that can be identified as the author's name from the input html file, text file, text, or the like. There are various character strings that can be specified at this time. For example, a character string such as “name, contributor, by, nickname” or a character string that identifies the writer is described in a set. is there.

  Step 102) The multi-author inspection processing unit 500 inspects whether or not there is a date before and after the name specified in Step 101.

  For example, when the date indicates December 20, 2004, there are various description methods such as “2004.12.20, 2004/12/20, 04.12.20, 2004 12 20, Dec. 20, 2004”. In addition, a date may be described together with a character string such as “date, posting date, question date”, or a date may be described along with a time. However, since the date is not necessarily required, this step can be omitted.

  If there is a date before and after the name, or if the input is an html file as shown in FIG. 9, there is a method of checking the tag structure and adopting the one surrounded by the same tag as the name. At this time, in FIG. 9, the word-of-mouth communication set is delimited by html <table> tags. There is also a method of adopting the one closest to the name information (the one with the smaller total number of characters from the character string indicating the name to the character string indicating the date).

  Step 103) It is checked whether or not there is a body (= review set) composed of one or more sentences together with the name collected in Step 101 and the date collected in Step 102.

  If a description “name: Keiko” is found in step 101 and “post date: 2004.12.20” is found in step 102 as shown in FIG. 7, whether or not there is a text composed of one or more sentences. inspect. However, if step 102 is omitted, the word-of-mouth set is formed with two pieces of information, that is, the name collected in step 101 and a body text composed of one or more sentences. Therefore, the inspection here is to check whether or not there is the above-described two sets of information.

  Step 104) It is checked whether or not the order structure of the review set acquired in Step 103 appears repeatedly.

  If there is a body consisting of one or more sentences with the description “Name:” and “Posted Date: yyyy.mm.dd” as shown in FIG. 7, is the structure of the review set repeated two or more times? Inspect whether. Here, yyyy represents the year (example: 2004), mm represents the month (example: 12), and dd represents the day (example: 20). However, if the processing in step 102 is omitted, it may be checked whether there are two or more reviews set repetitions excluding “post date: yyyy.mm, dd” indicating the date.

  If there are multiple groups of one or more sentences that can be regarded as the body text, for example, if the input is an html file as shown in Fig. 9, check the tag structure and separate it with the same tag as the name and date There is a method of adopting a set of sentences as text. In addition, there is a method of adopting a group of sentences closest to the name or date, or a method of adopting a candidate of a group of sentences having the longest sentence as the text.

  Step 105) If the review set appears repeatedly in Step 104, the number of repetitions of the review set acquired in Step 104 is calculated. Here, each data of the review set and the number of review sets are stored in the data storage unit 501.

  Step 106) Each name of the word-of-mouth communication set stored in the data storage unit 501 is compared, and it is checked whether or not there are a plurality of types. In the case of FIG. 7, the names correspond to “Keiko”, “I love hot springs”, and “Secret”. At this time, since all three are not the same, it is determined that “a plurality of names exist”.

  That is, it is determined whether or not the input is written by a plurality of writers. This determination is determined to be a multiple writer when word-of-mouth sets appear repeatedly and there are two or more names. This step 106 can be omitted. However, if step 106 is omitted, if the word-of-mouth set has repeatedly appeared, it is determined that there are multiple authors. At this time, the output information (whether or not it is a plurality of authors, and if it is a plurality of authors, each word-of-mouth communication set and the number thereof) is stored in the data storage unit 501.

  In the above, an example of starting from a name is shown, but even when starting from C “date”, the order of steps 101 and 102 described above is switched. The first step is to look for the date, and then look for the name around the date. However, when a review set in which C is omitted is used, processing starting from C cannot be applied.

  Next, the text extraction process in step 200 in the text extraction unit 510 will be described.

  FIG. 11 is a flowchart of text extraction processing according to the first embodiment of this invention.

  Step 201) The text extracting unit 510 reads each text of the word-of-mouth set and the number of texts S (S> 0) stored in the data storage unit 501 acquired by the multi-author inspection processing unit 500, and extracts HTML tags and the like. Determine if there is a symbol. If not, the process proceeds to step 203.

  Step 202) For each body, unnecessary symbols such as HTML tags other than sentences are removed.

  Step 203) Substitute 1 for s. Here, s is a text count.

  Step 204) If s is larger than S, the process is terminated.

  Step 205) The sth body is stored in the data storage unit 501.

  Step 206) Add 1 to s and proceed to Step 204.

  As described above, the text extracting unit 510 extracts a group of sentences determined to be text and stores them in the data storage unit 501. The text storage / output format is not limited to the above example, and can be selected, but is executed in a unified format.

  If no HTML tag is included in step 201, this processing is not necessary.

  Next, the evaluation expression inspection process in step 300 in the evaluation expression inspection processing unit 520 will be described.

  FIG. 12 is a flowchart of the evaluation expression checking process according to the first embodiment of this invention. In the following processing, it is assumed that morphological analysis processing is performed in advance.

  Step 301) The text of each word-of-mouth set and the number S of text are input from the data storage unit 501.

  Step 302) The evaluation expression inspection processing unit 520 substitutes 1 for s.

  Step 303) If s> S (the number of texts), the process ends.

  Step 304) The evaluation expression inspection processing unit 520 performs morphological analysis on the sth body. In morphological analysis, specific parts of speech such as nouns, adjectives, adjective verbs, and conjunctions that correspond to evaluation expressions are extracted.

  Step 305) The morphological analysis result and the evaluation expression stored in the evaluation expression storage unit 20 are collated to determine whether there is a matching expression. When the part of speech to be used is to be inspected, it is more efficient to inspect it after returning the written phrase to its original form.

  Step 306) If there is a matching expression, the process proceeds to step 307; otherwise, the process proceeds to step 308.

  Step 307) It is determined that “the s-th body has an evaluation expression”, the determination result (with an evaluation expression) is stored in the data storage unit 501, and the process proceeds to Step 309.

  Step 308) It is determined that “the s-th body has no evaluation expression”, the determination result (no evaluation expression) is stored in the data storage unit 501, and the process proceeds to Step 309.

  Step 309) Add 1 to s and proceed to Step 303.

  As a result of the above processing, what is output to the data storage unit 501 is the presence or absence of an evaluation expression in each text.

  In the above, an example in which it is determined in advance that the morpheme analysis process is performed is shown. However, if the morpheme analysis process is not performed, the sth is used instead of step 304 and step 305. The evaluation expression stored in the evaluation expression storage unit 20 is collated by character string matching to check whether there is a matching expression.

  Next, the experience expression inspection process in step 400 in the experience expression inspection processing unit 530 will be described.

  FIG. 13 is a flowchart of the experience expression inspection process in the first embodiment of the present invention. In the following processing, it is assumed that morphological analysis processing is performed in advance.

  Step 401) The experience expression inspection processing unit 530 receives the text of each word-of-mouth set and the number S of texts from the data storage unit 501.

  Step 402) Substitute 1 into s.

  Step 403) If s> S, the process is terminated.

  Step 404) Perform morphological analysis on the sth body. In morphological analysis, it is desirable to extract specific parts of speech such as nouns, adjectives and adjective verbs that correspond to the experiential expression.

  Step 405) The morphological analysis result and the experience expression stored in the experience expression storage unit 21 are collated to check whether there is a matching expression. When the part of speech to be used is to be inspected, it is more efficient to inspect it after returning the written words to the original form. Also, since the experiential expression is often the past form of the verb, there are also ways to focus on the verbs that come with the final forms of “ta” and “shim”, such as “I went and tried” is there.

  Step 406) If there is a matching expression, the process proceeds to step 407; otherwise, the process proceeds to step 408.

  Step 407) It is determined that “the s-th body has experience expression”, and the result (with experience expression) is stored in the data storage unit 501, and the process proceeds to Step 409.

  Step 408) It is determined that “the s-th body has no experience expression”, and the result (no experience expression) is stored in the data storage unit 501.

  Step 409) Add 1 to s, and go to Step 403.

  What is output to the data storage unit 501 in the above processing is the presence or absence of experience expression in each text.

  In the above, an example in which it is determined in advance that the morpheme analysis process is performed is shown. However, if the morpheme analysis process is not performed, the sth is used instead of step 404 and step 405. And the evaluation expression stored in the experience expression storage unit 21 are matched by character string matching to check whether there is a matching expression.

  Note that the evaluation expression test process (step 300) and the experience expression test process (step 400) are not related to each other, and can be performed at the same time or can be executed first.

  Further, when both the morphological analysis is performed by the evaluation expression inspection processing unit 520 and the experience expression inspection processing unit 530, it is desirable to perform the morpheme analysis before executing both processes in order to speed up the processing.

  An example of the series of flows shown in FIGS. 12 and 13 will be described.

FIG. 14 shows an example of a morphological analysis result in the first embodiment of the present invention. The example of the morphological analysis result shown in the figure is the text of the first review set of the three review sets in the review information input example of FIG. 7 “I was also very good yesterday…”
, "Takeshi Tsuji et al.," Japanese Morphological Analysis System Considering Conservativeness ", Report of the Society for Natural Language Processing, Information Processing Society of Japan, 117-09, vol.1997-no.004, pp.59-66 The results of morphological analysis using the morphological analysis system listed in Fig. 5 are shown. In this result example, “words extracted” by morphological analysis, “parts of speech” in which the words are classified, “reading” of the words, and “original form” of the words are shown separated by spaces. For example, “I” is classified as a part of speech called “pronoun of noun”, and it reads “I” and the original form is “I”.

  Thereafter, the evaluation expression inspection processing unit 520 collates the word output as the original form with the evaluation expression stored in the evaluation expression storage unit 20. If there is an example of the evaluation expression storage unit 20 as shown in FIG. 14, “good” matches.

  In the experience expression inspection processing unit 530, the experience expression stored in the experience expression storage unit 21 is collated with the morphological analysis result. When there is an example of the experience expression storage unit 21 as shown in FIG. 14, attention is paid to a word that has been morphologically analyzed as a verb because the past form of the verb is stored in the experience expression storage unit 21. In the morphological analysis result example of FIG. 14, the verb is related to “verb stem, verb inflection ending, verb suffix, verb stem continuation, verb suffix ending”. “I went” is a past tense, and since it is a polite expression of “I went”, it is considered the same, so it is stored in the example of the experience expression storage unit 21. Matches “I went”.

  Next, the word-of-mouth information determination processing in step 500 in the word-of-mouth information determination processing unit 540 will be described.

  FIG. 15 is a flowchart of word-of-mouth information determination processing according to the first embodiment of this invention.

  Step 501) The word-of-mouth information determination processing unit 540 reads from the data storage unit 501 the number S of each body text, presence / absence of experience expressions included in each body text, and presence / absence of evaluation expressions included in each body text. Here, it is assumed that data as shown in FIG. 16 has been read.

  Step 502) Substitute 1 for s and 0 for k. Here, s is a count of the number of documents, and k is the number of information sets determined to be word-of-mouth expressions.

  Step 503) If s> S, go to Step 507.

  Step 504) It is determined whether there is a word of mouth expression in the sth body.

  Step 505) If there is a word-of-mouth expression in step 504, 1 is added to k.

  Step 506) Add 1 to s, and go to Step 503.

  Step 507) If k / S ≧ X (X is a preset threshold value), the process proceeds to Step 508, and otherwise, the process proceeds to Step 509.

  Step 508) It is determined that the input document is word-of-mouth information.

  Step 509) It is determined that the input document is not word-of-mouth information.

  Next, the processing in step 504 will be described in detail.

  FIG. 17 is a flowchart for determining whether or not there is a word-of-mouth expression according to the first embodiment of this invention.

  Step 5041) It is determined whether there is both an evaluation expression and an experience expression in the s-th body read from the data storage unit 501 (for example, the example of FIG. 16). In the case, the process proceeds to step 5043.

  Step 5042) It is determined that “there is a word of mouth” in the sth body.

  Step 5043) It is determined that there is no word-of-mouth expression in the sth body.

  Note that the multi-writer inspection process shown in FIG. 10 and the text extraction process shown in FIG. 11 in the first embodiment may be omitted.

[Second Embodiment]
Also in the present embodiment, as in the first embodiment, an example is shown in which it is determined whether or not the input document is word-of-mouth information and whether or not there is a word-of-mouth expression in the sth body.

  In the present embodiment, the multi-writer inspection process is the same operation as that of the first embodiment described above, and thus the description thereof is omitted. It should be noted that whether or not to perform the multi-writer inspection process is arbitrary and can be omitted.

  FIG. 18 is a flowchart of text extraction in the second embodiment of this invention. In the figure, the same operations as those in FIG. 11 are denoted by the same step numbers, and the description thereof is omitted.

  In step 204, it is determined whether s> S. If s ≦ S, the number of sentences N (s) included in the sth body is detected, and N (s) is stored in the data storage unit 501. (N (s)> 0) (step 221). The subsequent processing is the same as in FIG. However, what is output to the data storage unit 501 is the number of sentences N (1),..., N (S) included in each text in addition to the text of each word-of-mouth set.

  Next, the evaluation expression inspection process in this embodiment will be described.

  FIG. 19 is a flowchart of the evaluation expression checking process in the second embodiment of the present invention.

  Step 321) The text of each word-of-mouth set and the number S of text are input from the data storage unit 501 to the evaluation expression inspection processing unit 520.

  Step 322) The evaluation expression inspection processing unit 520 substitutes 1 for s.

  Step 323) If s> S, the process ends.

  Step 324) Perform morphological analysis on the sth body.

  Step 325) The morpheme analysis result and the evaluation expression stored in the evaluation expression storage unit 20 are collated, and the matching expression and the number h (s) of the matching expressions are examined.

  Step 326) If there is a matching expression in the evaluation expression storage unit 20 in Step 325 and the number of matching expressions h (s)> 0, the process proceeds to Step 327. Otherwise, the process proceeds to Step 328. Transition.

  Step 327) It is determined that “the s-th body has an evaluation expression”, the matched expression and the number h (s) of the matched expressions are stored in the data storage unit 501, and the process proceeds to Step 329.

  Note that storing matching expressions in the data storage unit 501 is not essential, and is necessary when outputting, or when adding a weight to a specific expression.

  Step 328) It is determined that “the s-th body has no evaluation expression”, and the determination result (no evaluation expression) is stored in the data storage unit 501.

  Step 329) Add 1 to s and proceed to Step 323.

  In the above processing, the data storage unit 501 outputs the presence / absence of evaluation expressions in each text, the evaluation expressions in the text with evaluation expressions, and the number h of evaluation expressions in the text with evaluation expressions h ( s).

  In the above, an example in which it is determined in advance that the morpheme analysis process is performed is shown. However, if the morpheme analysis process is not performed, the s-th is used instead of steps 324 and 325. The evaluation expression stored in the evaluation expression storage unit 20 is collated by character string matching to check whether there is a matching expression.

  FIG. 20 is a flowchart of the experience expression inspection process in the second embodiment of the present invention.

  Step 421) The text of each word-of-mouth set and the number S of text are input from the data storage unit 501 to the experience expression test processing unit 530.

  Step 422) Substitute 1 for s.

  Step 423) If s> S, the process is terminated.

  Step 424) Morphological analysis is performed on the sth body. In morphological analysis, it is desirable to extract specific parts of speech such as nouns, adjectives and adjective verbs that correspond to the experiential expression.

  Step 425) The morphological analysis result and the experience expression stored in the experience expression storage unit 21 are collated, and the number of matching expressions and the number of matching expressions are checked.

  Step 426) If there is a matching expression (t (s)> 0), the process proceeds to step 427, and if not, the process proceeds to step 428.

  Step 427) It is determined that “the s-th body has an experience expression”, the number of matched expressions t (s) (with the matched expression) is stored in the data storage unit 501, and the process proceeds to Step 409. Further, as data stored in the data storage unit 501, in addition to “with experience expression”, experience expressions and the number t (s) thereof are accumulated.

  Note that storing the number of matching expressions in the data storage unit 501 is not essential, and is necessary when outputting, or when adding a weight to a specific expression.

  Step 428) It is determined that “the s-th body has no experience expression” and is stored in the data storage unit 501.

  Step 429) Add 1 to s and proceed to Step 423.

  In this processing, as shown in FIG. 21, the data storage unit 501 outputs presence / absence of experience expression in each text, experience expression in the text with experience expression, and experience in the text with experience expression. The number of expressions t (s).

  In the above, an example in which it is determined in advance that the morpheme analysis process is performed is shown. However, if the morpheme analysis process is not performed, the sth is used instead of steps 424 and 425. The evaluation expression stored in the experience expression storage unit 21 is collated by character string matching, and if there is a matching expression or if there is a matching expression, the number t (s) is checked.

  When both the evaluation expression inspection processing unit 520 and the experience expression inspection processing unit 530 perform morphological analysis, it is desirable to perform morphological analysis before executing both processes in order to speed up the processing.

  FIG. 22 is a flowchart of the word-of-mouth information determination process in the second embodiment of the present invention.

  The review determination process is the same for the processes other than step 501 and step 504 in the operation of FIG. 15 in the first embodiment.

In step 501, from the data storage unit 501, the number of texts: S;
・ The number of sentences included in each text: N (1), ..., N (S);
・ Experience expressions included in each text and the number thereof: t (1), ..., t (S);
-Evaluation expressions included in each text and the number thereof: h (1), ..., h (S);
Is entered.

  Hereinafter, the processing in step 504 in FIG. 22 in the present embodiment will be described.

  FIG. 23 is a flowchart for determining whether or not there is a word-of-mouth expression in the second embodiment of the present invention.

  Step 5051) The word-of-mouth information determination processing unit 540 determines whether there are both an evaluation expression and an experience expression in the sth body read from the data storage unit 501. In the case, the process proceeds to step 5043.

  Step 5052) The number of appearances of evaluation expressions h (s) included in each body and the number of appearances of experience expressions t (s) are added, and the number is the number of sentences N (s) included in each body of each review set. ) Is determined whether or not the number divided by a predetermined number Y is equal to or greater than ((h (s) + t (s)) / N (s) ≧ Y). If so, the process proceeds to step 5053, and if not, the process proceeds to step 5054.

  Step 5053) It is determined that “the word-of-mouth expression is present” in the sth body.

  Step 5054) It is determined that there is no word-of-mouth expression in the sth body.

  FIG. 24 shows an output example in the present embodiment.

In the present embodiment,
Number of text S;
the number N (s) of sentences contained in the sth body;
Experience expression included in the sth body and its number t (s);
Evaluation expressions included in the sth body and their number h (s);
When
The formula f (H, T) for determining whether or not there is a wording expression can also be calculated by weighting the entire evaluation expression and experience expression.

For example, when the weight of evaluation expression is E and the weight of experience expression is G,
f (H, T) = (E × h (s) + G × t (s)) / N (s) Equation (1)
It becomes.

  It is also possible to calculate by weighting only a specific word, not the entire expression.

  Since the number of appearances of each experience expression or evaluation expression is detected by the evaluation expression inspection processing unit 520 and the experience expression inspection processing unit 530, the following expression can be obtained.

Appearance types of evaluation expressions or the total number of evaluation expressions in the evaluation expression storage unit 20: AH;
Appearance types of experience expressions or total number of experience expressions in the experience expression storage unit 21: AT
A specific evaluation expression: nH (0 <n <AH);
A specific experience expression: mT (0 <m <AT);
the number of nH appearing in the sth body: nH (s) (0 ≦ nH (s) ≦ h (s));
number of mts appearing in the sth body: mT (s) (0 ≦ mT (s) ≦ t (s));
The weight on a particular expression nH: ne;
Weight on a particular expression nT: ng;
When
f (H, T) = Σ (ne × nH (s)) + Σ (ng × mT (s)) Equation (2)
If the weight for each representation is 1,
h (s) = ΣnH (s),
t (s) = ΣmT (s)
It becomes. That is, it is equal to the number of evaluation expressions and experience expressions included in the sth body.

Also, when the number of sentences included in the sth body text N (s),
f (H, T) = {Σ (ne × nH (s)) + Σ (ng × mT (s))} ÷ N (s) Equation (3)
It can also be expressed as If the expressions (1), (2), and (3) are f (H, T) ≧ Y (Y is a preset threshold value), it can be determined that the input is a word-of-mouth expression.

[Third Embodiment]
Also in the present embodiment, as in the first embodiment, an example is shown in which it is determined whether or not the input document is word-of-mouth information and whether or not there is a word-of-mouth expression in the sth body.

  In the present embodiment, the multi-writer inspection process is the same operation as that of the first embodiment described above, and thus the description thereof is omitted. It should be noted that whether or not to perform the multi-writer inspection process is arbitrary and can be omitted.

  FIG. 25 is a flowchart of text extraction in the third embodiment of the present invention. In the figure, the same operations as those in FIG. 11 are denoted by the same step numbers, and the description thereof is omitted.

  The processing from step 201 to step 204 is the same as that in FIG.

  Step 211) The text extracting unit 510 detects the number N (s) of sentences included in the sth text, and stores N (s) in the data storage unit 501 (N (s)> 0).

  Step 212) Substitute 1 for z. Where z is the count of sentences in the text.

  Step 213) It is determined whether z> N (s). If so, the process proceeds to Step 2114. If not, the process proceeds to Step 215.

  Step 214) Add 1 to s and proceed to Step 204.

  Step 215) The number n (s) (z) of sentences included in the zth sentence of the sth body is detected, and n (s) (z) is stored in the data storage unit 501 (n (s) ( z)> 0).

  Step 216) The z-th sentence of the s-th body is stored in the data storage unit 501.

  Step 217) Add 1 to z and go to Step 213.

  Through the operation shown in the flowchart, the contents output to the data storage unit 501 include the texts included in the texts of each review set, the number of sentences included in the texts of each review set N (1),. (S) and the number n (1) (1),..., N (S) (N (S)) of sentences included in each of the sentences included in the body text of each review set.

  Next, the evaluation expression inspection process in this embodiment will be described.

  FIG. 26 is a flowchart of the evaluation expression checking process in the third embodiment of the present invention.

  Step 341) The evaluation expression inspection processing unit 520 receives the number S of texts, the texts included in the texts, and the text numbers N (1), ..., N (S).

  Step 342) Substitute 1 for s and 0 for z. Here, s is the number of texts and z is the number of sentences.

  Step 343) If s is larger than S, the process is terminated.

  Step 344) Add 1 to z.

  Step 345) If z is larger than N (s), the process proceeds to Step 351.

  Step 346) Morphological analysis is performed on the zth sentence of the sth body.

  Step 347) The morpheme analysis result and the evaluation expression stored in the evaluation expression storage unit 20 are collated, and the number h (s) (z) of expressions that match the matching expressions is checked.

  Step 348) It is determined whether there is a matching expression (h (s) (z)> 0). If there is, the process proceeds to Step 349, and if not, the process proceeds to Step 350.

  Step 349) The matching expression and the number of expressions h (s) (z) are stored in the data storage unit 501, and the process proceeds to Step 344. Note that it is not essential to store the matching expressions and the number of expressions in the data storage unit 501, but it is necessary when outputting, or when adding a weight to a specific expression.

  Step 350) Store the number of matching expressions = 0 in the data storage unit 501, and proceed to Step 344.

  Step 351) If at least one evaluation expression is included in the sentence included in the sth body, it is determined that there is “evaluation expression”, and if it is not included, it is determined that there is no evaluation expression.

  Step 352) Add 1 to s, substitute 0 for z, and go to Step 343.

  As a result of the above processing, the data storage unit 501 outputs the presence / absence of evaluation expressions in each text, the evaluation expressions included in each text in the text with the evaluation expressions, and each text in the text having the evaluation expressions. Is the number h (s) (z) of evaluation expressions included in.

  In the above, an example in which it is determined in advance that the morpheme analysis process is performed is shown. However, if the morpheme analysis process is not performed, the sth is used instead of step 346 and step 347. The number of evaluation expressions h (s) (z) included in each sentence in the body with the matching expression and the evaluation expression. )

  FIG. 27 is a flowchart of the experience expression inspection process in the third embodiment of the present invention.

  Step 441) The experience expression inspection processing unit 530 receives from the data storage unit 501 the number S of texts, the texts included in the texts, and the text numbers N (1),..., N (S).

  Step 442) 1 is substituted into s, and 0 is substituted into z. Here, s is a text count, and z is a text count.

  Step 443) It is determined whether or not s> S. If so, the process ends.

  Step 444) Add 1 to z.

  Step 445) It is determined whether z> N (s). If so, the process proceeds to Step 451.

  Step 446) Perform morphological analysis on the zth sentence of the sth body.

  Step 447) The morpheme analysis result and the experience expression stored in the experience expression storage unit 21 are collated, and the matching expression and the number of matching expressions t (s) (z) are checked.

  Step 448) It is determined whether there is a matching expression (t (s) (z)> 0). If there is a matching expression, the process proceeds to Step 449, and if not, the process proceeds to Step 450.

  Step 449) The number t (s) (z) of the matching expressions and the matching expressions are stored in the data storage unit 501, and the process proceeds to Step 444. Note that it is not essential to store the number of matching expressions with the matching expressions in the data storage unit 501. This is necessary when outputting, or when adding a weight to a specific expression.

  Step 450) The number of matching expressions (t (s) (z) = 0) is stored in the data storage unit 501, and the process proceeds to Step 444.

  Step 451) If at least one experiential expression is included in the text included in the sth body, “experience expression is present”, the experiential expression and the number of experiential expressions are stored in the data storage unit 501 and included. Otherwise, it is determined that “no experience expression” and “no experience” is stored in the data storage unit 501.

  Step 452) Add 1 to s and substitute 0 for z.

  As a result of the above processing, the data storage unit 501 includes the presence / absence of experience expression in each text, the experience expression included in each sentence in the text with the experience expression, and each sentence in the text with the experience expression. The number of experience expressions t (s) (z) to be stored is stored.

  In the above, an example in which it is determined in advance that the morpheme analysis process is performed is shown. However, if the morpheme analysis process is not performed, the sth is used instead of step 446 and step 447. The number of experience expressions t (s) (z) included in each sentence in the body with the matching expression and the experience expression are matched by character string matching. )

  Next, word-of-mouth information determination processing is performed. The review determination process is the same for the processes other than step 501 and step 504 of FIG. 15 in the first embodiment described above.

FIG. 28 is a flowchart of the word-of-mouth information determination process in the third embodiment of the present invention. In step 501,
-Number of text S;
・ The number of sentences included in each text N (1), ..., N (S);
The number of sentences included in each sentence of each body text n (1) (1),..., N (S) (N (S));
・ Experience expressions and their numbers t (1) (1),..., T (S) (N (S)) included in each sentence of each text;
・ Evaluation expressions included in each sentence of each body and the number h (1) (1),..., H (S) (N (S));
Is entered.

  The processing in step 504 in FIG. 15 in the present embodiment will be described below with reference to FIG.

  FIG. 29 is a flowchart of the determination of the presence / absence of word-of-mouth expression in the third embodiment of the present invention.

  In the present embodiment, it is assumed that the above-described contents are input from the data storage unit 501 to the word-of-mouth information determination processing unit 540 in step 501 of FIG. For example, the contents shown in FIG. 30 are input as the above example.

  Hereinafter, the processing of step 504 will be described.

  Step 5061) The word-of-mouth information determination processing unit 540 determines whether there are both evaluation expression and experience expression in the sth body. If there is, the process proceeds to Step 5062, and if not, the process proceeds to Step 5067. .

  Step 5062) Substitute 1 for z.

  Step 5063) It is determined whether z> N (s). If so, the process goes to Step 5067, and if not, the process goes to Step 5064.

  Step 5064) It is determined whether (h (s) (z) + t (s) (z)) / n (s) (z) ≧ Y (where Y is a preset threshold value). If not, the process proceeds to step 5065. Otherwise, the process proceeds to step 5066.

  In addition, instead of the above condition (h (s) (z) + t (s) (z)) / n (s) (z) ≧ Y, h (s) (z) + t (s) (z) It may be determined using ≧ Y.

  Step 5065) It is determined that “the word-of-mouth expression is present” in the sth body.

  Step 5067) It is determined that the s-th body has no word-of-mouth expression.

  With the above processing, the output as shown in FIG. 24 is obtained.

In the present embodiment,
・ Number of text: S;
・ The number of sentences included in each text: N (1), ..., N (S);
-Number of sentences included in each sentence of each body: n (1) (1), ..., n (S) (N (S));
・ Experience expressions included in each sentence of each body and the number thereof: t (1) (1),..., T (S) (N (S));
-Evaluation expressions included in each sentence of each body and the number thereof: h (1) (1), ..., h (S) (N (S));
When
f (H, T) can also be calculated by weighting the entire evaluation expression and experience expression.

For example, when the weight of evaluation expression is E and the weight of experience expression is G,
f (H, T) = (E × h (s) (z) + G × t (s) (z)) / n (s) (z) Equation (4)
It is also possible to calculate by weighting only a specific phrase, not the entire expression. Since the number of appearances of each experience expression or evaluation expression is detected by the evaluation expression inspection processing unit 520 and the experience expression inspection processing unit 530, the following expression can be obtained.

Appearance types of evaluation expressions or the total number of evaluation expressions in the evaluation expression storage unit 20: AH;
Type of appearance of experience expression or total number of experience expressions in experience expression storage unit 21: AT;
A specific evaluation expression: nH (0 <n <AH);
A specific experience expression: mT (0 <m <AT);
The number of nH appearing in the sentence z with the sth body: nH (s) (z)
(0 ≦ nH (s) (z) ≦ h (s) (z))
The number of mTs that appear in the sentence z in the sth body: mT (s) (z)
(0 ≦ mT (s) (z) ≦ t (s) (z))
The weight on a particular expression nH: ne;
Weight on a particular expression nT: ng;
When
f (H, T) = Σ (ne × nH (s) (z)) + Σ (ng × mT (s) (z)) Equation (5)
If the weight for each representation is 1,
h (s) (z) = ΣnH (s) (z),
t (s) (z) = ΣmT (s) (z)
It becomes. That is, it is equal to the number of evaluation expressions and experience expressions included in the zth sentence of the sth body.

Also,
f (H, T) = {Σ (ne × nH (s) (z)) + Σ (ng × mT (s) (z))}
÷ n (s) (z) Equation (6)
It can also be expressed as If both formulas (4), (5), and (6) are f (H, T) ≧ Y (Y is a preset threshold value), it can be determined that the input is a word-of-mouth expression.

  This f (H, T) is not limited to the above example, and can be expressed by a calculation formula using H and T.

[Fourth Embodiment]
This embodiment is different from the first embodiment only in word-of-mouth information determination processing. In the word-of-mouth information determination process in the present embodiment, the word-of-mouth rate of the input document is obtained. Since other processes are the same as those in the first embodiment, only the word-of-mouth information determination process will be described below.

  FIG. 31 is a flowchart of the word-of-mouth information determination process in the fourth embodiment of the present invention.

Step 701) The following information is input from the data storage unit 501 to the review information determination processing unit 540.
・ The number S of each text;
・ With / without experience expression included in each text;
・ With / without evaluation expression included in each text;
Step 702) Substitute 1 for s and 0 for k. Here, s is a text count, and k is the number of word-of-mouth information sets determined to be word-of-mouth information.

  Step 703) If s> S, go to Step 707.

  Step 704) It is determined whether there is a word-of-mouth expression in the s-th body. For the detailed processing of this step, the above-described processing of FIG. 17 is performed.

  Step 705) Add 1 to k.

  Step 706) Add 1 to s, and go to Step 703.

  Step 707) The word-of-mouth rate of the input document is calculated by K / S * 100 (%) and output. However, the above k and K are the number of information sets determined as word-of-mouth expressions.

  The above step 707 will be described. In step 707, it is determined whether the entire input document is word-of-mouth information or the word-of-mouth rate is what percentage. This determination method can be expressed by the following function.

f (K, S)
K: Number of word-of-mouth information sets determined as word-of-mouth expression ≧ 0
S: Number of input review information sets (= number of texts) ≧ 1
For example,
f (K, S) = K ÷ S,
f (K, S) can be expressed as a word-of-mouth rate as follows.

f (K, S) = K ÷ S × 100
In this embodiment, only the word-of-mouth rate is output.

[Fifth Embodiment]
This embodiment is different from the second embodiment only in the word-of-mouth information determination process. In the word-of-mouth information determination process in the present embodiment, the word-of-mouth rate of the input document is obtained. Since other processes are the same as those in the second embodiment, only the word-of-mouth information determination process will be described below.

  FIG. 32 is a flowchart of word-of-mouth information determination processing in the fifth embodiment of the present invention.

In step 801,
・ Number of text: S;
・ The number of sentences included in each text: N (1), ..., N (S);
・ Experience expressions included in each text and the number thereof: t (1), ..., t (S);
-Evaluation expressions included in each text and the number thereof: h (1), ..., h (S);
Is entered.

  The review information determination process is the same as the process of FIG. 31 in the fourth embodiment described above. However, in step 704, an expression ((h (s) + t (s)) / N (s) ≧ Y) for determining whether or not there is a word-of-mouth expression shown in step ° 5052 in FIG.

  With the above processing, in the present embodiment, a result as shown in FIG. 33 is output.

  In the present embodiment, it is also possible to add a weight in the word-of-mouth determination process as in the second embodiment described above.

[Sixth Embodiment]
This embodiment is different from the above-described third embodiment only in that the word-of-mouth rate of the input document is obtained in the word-of-mouth information determination process.

  FIG. 34 is a flowchart of the word-of-mouth determination process in the sixth embodiment of the present invention. The flowchart shown in the figure is different from the flowchart of FIG. 31 in the above-described fourth embodiment only, and the other processes are the same as those of the fourth embodiment, and thus description thereof is omitted.

Step 901) The following information is input from the data storage unit 501 to the review information determination processing unit 540.
・ Number of texts: S;
・ The number of sentences included in each text: N (1), ..., N (S);
The number of sentences included in each sentence of each body text n (1) (1),..., N (S) (N (S));
・ Experience expressions included in each sentence of each body and the number thereof: t (1) (1),..., T (S) (N (S));
-Evaluation expressions included in each sentence of each body and the number thereof: h (1) (1), ..., h (S) (N (S));
Further, in the present embodiment, as the processing in step 704, the processing in FIG. 29 is performed in the same manner as in the third embodiment described above, and thus the description thereof is omitted.

  As a result of the above processing, the result shown in FIG. 33 is output.

  In the present embodiment, it is also possible to add a weight in the word-of-mouth determination process as in the third embodiment.

  The operations shown in the first to sixth embodiments are constructed as a program and executed by a computer having the evaluation expression storage unit 20, the experience expression storage unit 21, and the data storage unit 501, or via a network. Can be distributed.

  Also, the constructed program is stored in a hard disk device connected to a computer used as a word-of-mouth information determination device, a portable storage medium such as a flexible disk, a CD-ROM, etc., and installed in a computer or distributed. Is possible.

  The present invention is not limited to the above-described embodiment, and various modifications and applications can be made within the scope of the claims.

  The present invention can be applied to a technique for determining a user's voice and reputation regarding a certain thing, and a Web search classification technique.

It is a principle block diagram of this invention. It is a figure for demonstrating the principle of this invention. It is a figure for demonstrating the outline | summary in one embodiment of this invention. It is a block diagram of the review information determination apparatus in one embodiment of this invention. It is an example of the experience expression in one embodiment of this invention. It is a flowchart of the outline | summary operation | movement of the review information determination process in one embodiment of this invention. It is an example of word-of-mouth information input in one embodiment of the present invention. It is a figure which shows the relationship between the input document in one embodiment of this invention, the text, a sentence, and a sentence. It is an example of the word-of-mouth information input html structure in the 1st Embodiment of this invention. It is a flowchart (example which starts from a name) of the multi-writer test | inspection process in the 1st Embodiment of this invention. It is a flowchart of the text extraction process in the 1st Embodiment of this invention. It is a flowchart of the evaluation expression test | inspection process in the 1st Embodiment of this invention. It is a flowchart of the experience expression test | inspection process in the 1st Embodiment of this invention. It is an example of the morphological analysis result in the 1st Embodiment of this invention. It is a flowchart of the review information determination process in the 1st Embodiment of this invention. It is an example of data of the data storage part in the 1st Embodiment of this invention. It is a flowchart of the determination of the presence or absence of the word-of-mouth expression in the 1st Embodiment of this invention. It is a flowchart of the text extraction process in the 2nd Embodiment of this invention. It is a flowchart of the evaluation expression inspection process in the 2nd Embodiment of this invention. It is a flowchart of the experience expression test | inspection process in the 2nd Embodiment of this invention. It is a flowchart of the determination of the presence or absence of the word-of-mouth expression in the 2nd Embodiment of this invention. It is a flowchart of the review information determination process in the 2nd Embodiment of this invention. It is a flowchart of the determination of the presence or absence of the word-of-mouth expression in the 2nd Embodiment of this invention. It is an output example in the 2nd Embodiment of this invention. It is a flowchart of the text extraction process in the 3rd Embodiment of this invention. It is a flowchart of the evaluation expression inspection process in the 3rd Embodiment of this invention. It is a flowchart of the experience expression test | inspection process in the 3rd Embodiment of this invention. It is a flowchart of the review information determination process in the 3rd Embodiment of this invention. It is a flowchart of the determination of the word-of-mouth expression presence or absence in the 3rd Embodiment of this invention. It is an example of data of the data storage part in the 3rd Embodiment of this invention. It is a flowchart of the review information determination process in the 4th Embodiment of this invention. It is a flowchart of the review information determination process in the 5th Embodiment of this invention. It is an output example in the 5th Embodiment of this invention. It is a flowchart of the review information determination process in the 6th Embodiment of this invention.

Explanation of symbols

10 Computer 11 Display Unit 12 Instruction Input Unit 20 Evaluation Expression Storage Unit, Evaluation Expression Storage Unit 21 Experience Expression Storage Unit, Experience Expression Storage Unit 40 Network 501 Data Storage Unit, Data Storage Unit 500 Multi-Writer Examination Processing Unit 510 Text Extraction Processing Unit 520 Evaluation expression inspection means, evaluation expression inspection section 530 Experience expression inspection means, experience expression inspection section 540 Review information determination means, Review information determination section

Claims (10)

A word-of-mouth information determination device for determining word-of-mouth information including experiences and complaints about a certain thing from the input text,
An evaluation expression storage means for storing an evaluation expression, which is a phrase that evaluates the quality of things,
Experience expression storage means for storing experience expressions, which are phrases representing the use or experience / experience of things,
Referring to the evaluation expression storage means, it is determined whether or not an evaluation expression, which is a word that evaluates the quality of things, is described in the input text, and the determination result is stored in the data storage means. Evaluation expression inspection means;
Referring to the experience expression storage means, it is determined whether or not an experience expression that is a phrase representing use or experience / experience of a thing is described in the input text, and the determination result is a data storage means The experiential expression inspection means stored in the
When the information of the data storage means is read out, it is determined that the evaluation expression is described by the evaluation expression inspection means, and the experience expression is determined by the experience expression inspection means, A word-of-mouth determination means for determining whether or not a word-of-mouth is based on a word-of-mouth determination function represented by an evaluation expression and a calculation formula based on the experience expression;
The word-of-mouth information determination apparatus characterized by having. A word-of-mouth information determination device that determines word-of-mouth information including experiences and complaints about a certain thing in the text entered or the structure of a web page, and complaints,
An evaluation expression storage means for storing an evaluation expression, which is a phrase that evaluates the quality of things,
Experience expression storage means for storing experience expressions, which are phrases representing the use or experience / experience of things,
In the input text or the structure of the web page, a set of one or more sentences and a name or symbol including the writer name or nickname of the body, date information describing the body, Or a multi-writer inspection means for inspecting whether the text, the name or the symbol set repeatedly appears;
Referring to the evaluation expression storage means, it is determined whether or not an evaluation expression that is a phrase that evaluates the quality of things is described in the text, and the determination result is stored in the data storage means Evaluation expression inspection means to
With reference to the experience expression storage means, it is determined whether or not an experience expression that is a phrase representing use or experience / experience of a thing is described in the text, and the determination result is the data storage means The experiential expression inspection means stored in the
When the determination result is read from the data storage means, it is determined that the evaluation expression is described by the evaluation expression inspection means, and the experience expression is determined by the experience expression inspection means, A word-of-mouth determination means for determining whether or not the word-of-mouth is a word-of-mouth by a word-of-mouth determination function represented by a calculation formula based on the evaluation expression and the experience expression;
The word-of-mouth information determination apparatus characterized by having. A word-of-mouth information determination device for determining word-of-mouth information including experiences and complaints about a certain thing from the input text,
An evaluation expression storage means for storing an evaluation expression, which is a phrase that evaluates the quality of things,
Experience expression storage means for storing experience expressions, which are phrases representing the use or experience / experience of things,
Referring to the evaluation expression storage means, it is determined whether or not an evaluation expression, which is a word that evaluates the quality of things, is described in the input text, and the determination result is stored in the data storage means. Evaluation expression inspection means;
Referring to the experience expression storage means, it is determined whether or not an experience expression that is a phrase representing use or experience / experience of a thing is described in the input text, and the determination result is a data storage means The experiential expression inspection means stored in the
When the determination result is read from the data storage means, it is determined that the evaluation expression is described by the evaluation expression inspection means, and the experience expression is determined by the experience expression inspection means, A word-of-mouth rate calculation means for calculating a word-of-mouth rate, which is a ratio occupied by the word-of-mouth information, of the whole of the input text by a word-of-mouth determination function represented by a calculation formula based on the evaluation expression and the experience expression;
The word-of-mouth information determination apparatus characterized by having. A word-of-mouth information determination device that determines word-of-mouth information including experiences and complaints about a certain thing in the text entered or the structure of a web page, and complaints,
An evaluation expression storage means for storing an evaluation expression, which is a phrase that evaluates the quality of things,
Experience expression storage means for storing experience expressions, which are phrases representing the use or experience / experience of things,
In the input text or the structure of the web page, a set of one or more sentences and a name or symbol including the writer name or nickname of the body, date information describing the body, Or a multi-writer inspection means for inspecting whether the text, the name or the symbol set repeatedly appears;
Referring to the evaluation expression storage means, it is determined whether or not an evaluation expression that is a phrase that evaluates the quality of things is described in the text, and the determination result is stored in the data storage means Evaluation expression inspection means;
Referring to the experience expression storage means, it is determined whether or not an experience expression that is a phrase representing use or experience / experience of things is described in the text, and the determination result is stored in the data storage means Storing experiential expression inspection means;
When the determination result is read from the data storage means, it is determined that the evaluation expression is described by the evaluation expression inspection means, and the experience expression is determined by the experience expression inspection means The word-of-mouth rate calculation that calculates the word-of-mouth rate, which is the ratio occupied by the word-of-mouth information in the input text or the entire Web page, by the word-of-mouth determination function represented by the calculation expression based on the evaluation expression and the experience expression Means,
The word-of-mouth information determination apparatus characterized by having. In the word-of-mouth information determination method for determining word-of-mouth information including experiences related to certain things, comments used, and complaints from the entered text,
Referring to an evaluation expression storage means for storing an evaluation expression that is a phrase that evaluates the quality of things, it is determined whether or not the evaluation expression is described in the input text, and the determination result is stored in data An evaluation expression checking step stored in the means;
Referring to the experience expression storage means for storing the experience expression, which is a phrase representing the use or experience / experience of things, it is determined whether the experience expression is described in the input text, and the determination result is obtained. An experiential expression inspection step stored in the data storage means;
When the determination result is read from the data storage means, it is determined that the evaluation expression is described in the evaluation expression inspection step, and the experience expression is determined to be described in the experience expression inspection step A word-of-mouth determination step for determining whether or not the word-of-mouth is a word-of-mouth by a word-of-mouth determination function represented by a calculation formula based on the evaluation expression and the experience expression;
A word-of-mouth information determination method characterized by: In the word-of-mouth information determination method for determining word-of-mouth information including experiences, complaints, and complaints about certain things from the text entered or the structure of the web page,
In the input text or the structure of the web page, a set of one or more sentences and a name or symbol including the writer name or nickname of the body, date information describing the body, Or a multi-writer inspection step for inspecting whether the text, the name or the symbol pair appears repeatedly;
Referring to an evaluation expression storage means for storing an evaluation expression that is a phrase that evaluates the quality of things, it is determined whether or not the evaluation expression is described in the text, and the determination result is data An evaluation expression checking step to be stored in the storage means;
Referring to experience expression storage means for storing experience expressions, which are phrases representing the use or experience / experience of things, it is determined whether the experience expression is described in the text, and the determination result is data An experiential expression inspection step stored in a storage means;
When the determination result is read from the data storage means, it is determined that the evaluation expression is described in the evaluation expression inspection step, and the experience expression is determined to be described in the experience expression inspection step A word-of-mouth determination step for determining whether or not the word-of-mouth is a word-of-mouth by a word-of-mouth determination function represented by a calculation formula based on the evaluation expression and the experience expression;
A word-of-mouth information determination method characterized by: In the word-of-mouth information determination method for determining word-of-mouth information including experiences related to certain things, comments used, and complaints from the entered text,
Referring to an evaluation expression storage means for storing an evaluation expression that is a phrase that evaluates the quality of things, it is determined whether or not the evaluation expression is described in the input text, and the determination result is stored in data An evaluation expression checking step stored in the means;
Referring to the experience expression storage means for storing the experience expression, which is a phrase representing the use or experience / experience of things, it is determined whether the experience expression is described in the input text, and the determination result is obtained. An experiential expression inspection step stored in the data storage means;
When the determination result is read from the data storage means, it is determined that the evaluation expression is described in the evaluation expression inspection step, and the experience expression is determined to be described in the experience expression inspection step A word-of-mouth rate calculation step of calculating a word-of-mouth rate, which is a ratio occupied by the word-of-mouth information in the whole of the input text, by a word-of-mouth determination function represented by a calculation formula based on the evaluation expression and the experience expression;
A word-of-mouth information determination method characterized by: In the word-of-mouth information determination method for determining word-of-mouth information including experiences, complaints, and complaints about certain things from the text entered or the structure of the web page,
In the input text or the structure of the web page, a set of one or more sentences and a name or symbol including the writer name or nickname of the body, date information describing the body, Or a multi-writer inspection step for inspecting whether the text, the name or the symbol pair appears repeatedly;
Referring to an evaluation expression storage means for storing an evaluation expression that is a phrase that evaluates the quality of things, it is determined whether or not the evaluation expression is described in the text, and the determination result is data An evaluation expression checking step to be stored in the storage means;
Refers to experience expression storage means for storing experience expressions that are words representing use or experience / experience of things, determines whether the experience expressions are described in the text, and stores the determination results as data An experiential expression inspection step stored in the means;
When the determination result is read from the data storage means, it is determined that the evaluation expression is described in the evaluation expression inspection step, and the experience expression is determined to be described in the experience expression inspection step The word-of-mouth rate calculation that calculates the word-of-mouth rate, which is the ratio occupied by the word-of-mouth information in the input text or the entire Web page, by the word-of-mouth determination function represented by the calculation expression based on the evaluation expression and the experience expression Steps,
A word-of-mouth information determination method characterized by: A word-of-mouth information determination program for determining word-of-mouth information including experience stories, impressions used, and complaints from the text entered or the structure of a web page,
An evaluation expression storage means for storing an evaluation expression, which is a phrase that evaluates the quality of things,
In a computer equipped with experience expression storage means for storing experience expressions that are words representing use or experience / experience of things,
A word-of-mouth information determination program for executing the word-of-mouth information determination method according to any one of claims 5 to 8. A storage medium storing a word-of-mouth information determination program for determining word-of-mouth information including experiences, complaints, and complaints related to certain things from the input text or the structure of a web page,
An evaluation expression storage means for storing an evaluation expression, which is a phrase that evaluates the quality of things,
In a computer equipped with experience expression storage means for storing experience expressions that are words representing use or experience / experience of things,
9. A storage medium storing a word-of-mouth information determination program, wherein a program for executing a process for realizing the word-of-mouth information determination method according to claim 5 is stored.

Images