JP5404575B2 - EXPERIENCE FACTOR RELATIONSHIP APPARATUS AND METHOD AND PROGRAM - Google Patents

Description

  The present invention relates to an experience factor relationship analysis apparatus, method, and program, and in particular, experience factor relationship analysis using an experience article set with experience attributes for classifying text (articles) from the viewpoint of factors that have led to experience. The present invention relates to an apparatus, a method, and a program.

  Here, the experience article is an article in which people have described some experience, and the experience attribute is the experience type of the experience article (for example, “go to the hamburger shop” as the attribute of the experience article A). A factor refers to an event or condition (an event or condition prior to experience) that is a major cause of people's experience.

  The acquisition of causal knowledge is important in various fields such as economic forecasting, marketing, and behavioral navigation. As the types of causal relationship knowledge, there are causal relationships between events in the world and causal relationships between individual behaviors and their factors. As for the former, there is a conventional technique for automatically acquiring a large number of articles as an information source (see, for example, Non-Patent Document 1).

  The above-mentioned method such as Non-Patent Document 1 uses newspaper articles as an information source, and adapts to a specific language pattern for each article text using clues indicating connection signs indicating factors such as “for” and “so” Identify expressions as places where factors and results are written. Furthermore, for each expression of the factor and result, normalization is performed by word processing etc. for what each is about, the pair relation of the factor and result is extracted, and further the factor relation is collected by collecting a large amount of the pair relations. It is a method of acquiring the strength of the relationship between results as knowledge.

Inui Takashi, Inui Kentaro, Matsumoto Yuji. "Automatic acquisition of causal knowledge from a document set based on a connection sign", "IPSJ Journal, Vol.45, No.3, pp. 919-933, 2004)

  However, the conventional research that automatically acquires causal relations is effective for a set of articles such as newspaper articles with low ambiguity in determining whether or not the cause is a causal part, but an experience article written by a general user. I think that it may not work effectively for an article set with high ambiguity of expression. This is because there are many cases in which experience articles written by general users can represent factors without an explicit connection indicator.

  An example is shown below. In Example 1, the explicit connection expression “So” depends on the action expression “I went”.

  Example 1) I went to McDonald's Yokosuka Chuo store because I could use the power source without hesitation.

  On the other hand, in Example 2, there is no explicit connection expression, but there is a factor.

  Example 2) There are a lot of things around the outlet, and McDonald's is the choice.

Example 3) Go to the neighborhood Mac because I thought it would be great to be able to use the power supply without worrying about the surroundings !!
In the above-described conventional causal relationship extraction technique, it is assumed that the factors existing in the sentences of Example 2 and Example 3 cannot be extracted, and it is expected that this will cause a problem in appropriately classifying articles.

  As a result, we used a set of articles when the ambiguity of expressions such as experience articles written by general users when extracting the relationship between each person's behavior and its factors was used for marketing etc. When performing a causal relationship analysis, the conventional method cannot guarantee sufficient performance for acquiring knowledge between the factor and the result.

  The present invention has been made in view of the above points, and the strength of the relationship between people's experiences and factors, using an article set with high ambiguity in expression such as experience articles with experience attributes written by general users as an information source. It is an object of the present invention to provide an experience factor relationship analyzing apparatus, method, and program capable of calculating the relationship between experience and factors.

In order to solve the above-mentioned problems, the present invention (Claim 1) is a score representing the strength of the relationship between various experiences and various factors that are the main causes and states that people experience. Empirical factor relation analysis device that outputs
Experience article storage means with experience attributes storing experience articles that describe an experience article of people's own experience and an experience attribute that is an attribute indicating the type of experience of the experience article;
Experience attribute storage means storing experience attribute ID and experience attribute;
Factor attribute storage means storing factor attribute ID and factor attribute;
Experience article input means for reading an experience article set from the experience article storage means with experience attributes;
A word factor context score giving means for giving a word factor context score representing a degree of use of each word of each experience article in the experience article set in a context representing the factor;
For each word of each experience article to which a word factor context score has been assigned, the factor attribute storage means is referred to, and the experience article factor indicating what factor the experience described in each article is based on An experience article factor score giving means for giving a score;
Obtain an experience article set including the experience article factor score and experience attribute, obtain the experience article factor score, refer to the experience attribute storage means for each experience attribute of the experience article, and each experience and each factor Empirical factor relationship output means for outputting a total empirical factor relationship score indicating the strength of the relationship between.

In the present invention (Claim 2), in the word factor context score assigning means,
When giving the word factor context score to each word of each experience article,
Means for using the appearance position of each word in the document as a feature for assigning the word factor context score;

The present invention (Claim 3) provides the word factor context score assigning means,
When giving the word factor context score to each word of each experience article,
And means for using the appearance position relationship between each word and an experience expression (with a verb or a noun) representing experience as a feature for assigning the word factor context score.

Further, the present invention (Claim 4) is characterized in that in the word factor context score assigning means,
When giving the word factor context score to each word of each experience article,
And means for using a clue expression indicating that the factor is not written as a feature for assigning the word factor context score.

Further, according to the present invention (Claim 5), in the word factor context score assigning means,
When giving the word factor context score to each word of each experience article,
Means for using the appearance position of each word in the document as a feature for assigning the word factor context score;
Means for utilizing the position relationship between each word and an experience expression (with a verb or a noun) representing experience, as a feature that gives the word factor context score;
Means for using a clue expression indicating that the factor is not written as a feature for giving the word factor context score;
At least one means.

Further, the present invention (Claim 6) is the experience article factor score assigning means,
For each experience attribute, which is attributed to each experience attribute, the sum of the experience article factor scores of each experience article is calculated for each factor, and the value of each sum of the experience attributes and factors is calculated. A means for outputting the experience article factor score which is the strength of the relationship.

In the present invention (Claim 7), in the experience article factor score assigning means,
For each experience attribute that has it as an attribute, for each experience article set, the sum of experience article factor scores of each experience article for each factor (X1), the number of experience articles that have the experience as an attribute ( X2), the sum of the experience article factor scores (X3) and the total number of articles (X4) for the relevant factors of all experience articles, respectively. Includes means for outputting the relationship between the experience attribute and the factor as the total experience factor relationship score by an index that measures how much is higher than X3 compared to X3.

Further, the present invention (Claim 8) is an experiential factor relationship that outputs a score representing the strength of the relationship between various experiences and the various factors that are the main causes of people experiencing. An analysis method,
Experience article storage means with experience attributes storing experience articles that describe an experience article of people's own experience and an experience attribute that is an attribute indicating the type of experience of the experience article;
Experience attribute storage means storing experience attribute ID and experience attribute;
In an apparatus having factor attribute ID and factor attribute storage means storing factor attributes,
Experience article input means reads an experience article set from experience article storage means with experience attribute storage means,
A word factor context score assigning step in which the word factor context score assigning unit assigns a word factor context score indicating the degree of use of each word of each experience article in the experience article set in the context representing the factor. When,
The experience article factor score assigning means refers to the factor attribute storage means for each word of each experience article to which the word factor context score has been assigned, and the experience described in each article is based on what factor An experience article factor score granting step for giving an experience article factor score representing whether or not
The experience factor relation output means acquires an experience article set including the experience article factor score and experience attribute, acquires the experience article factor score, and refers to the experience attribute storage means for each experience attribute of the experience article Then, an experience factor relationship output step of outputting a total experience factor relationship score indicating the strength of the relationship between each experience and each factor is performed.

  Moreover, this invention (Claim 9) is a program for functioning a computer as each means which comprises the empirical factor relationship analysis apparatus of any one of Claim 1 thru | or 7.

  As described above, according to the present invention, the strength of the relationship between people's experiences and factors is calculated using a collection of articles with high ambiguity in expression, such as experience articles with experience attributes written by general users. It becomes possible to do.

  Knowing the strength of the relationship between experience and factors is useful in various fields, but it is particularly important in marketing. Realizing more effective marketing than ever is an example of the effect of the present invention. By using a device in which the present invention is implemented, marketers can know factors that are characteristic of a certain experience and experiences that are characteristic of a certain factor. It can be used to promote services and products.

It is a block diagram of the experience factor causal relationship analyzer in one embodiment of this invention. It is a structural example of experience article collection DB with experience attribute in one embodiment of this invention. It is a structural example of factor attribute DB in one embodiment of this invention. It is a structural example of experience attribute DB in one embodiment of this invention. It is a flowchart of the operation | movement in one embodiment of this invention. It is a block diagram of the word factor context score provision part in one embodiment of this invention. It is a figure which shows the parameter | index of each word in each document in one embodiment of this invention. It is an example (we assume the article regarding a restaurant visit) of the clue expression in one embodiment of the present invention. It is an example of the experience article factor score provided by the experience article factor score assigning unit in one embodiment of the present invention. It is an example of the experience factor score output in the experience factor relation score calculation part in one embodiment of this invention. It is an example of the article set (3 articles) which has the experience attribute A input into the experience factor relation cost calculation part in one embodiment of this invention. It is a figure for demonstrating the experience factor relation score of the experience factor relation score calculation part in one embodiment of this invention.

  Embodiments of the present invention will be described below with reference to the drawings.

  FIG. 1 shows the configuration of an empirical factor causal relationship analysis apparatus according to an embodiment of the present invention.

  The experience factor causal relationship analysis apparatus shown in the figure includes an experience article input unit 10, a word factor context score assignment unit 20, an experience article factor score assignment unit 30, an experience factor relation score calculation unit 40, an attributed experience article set DB 50, factors An attribute DB 60 and an experience attribute DB 70 are included.

  The experience article set DB 50 with experience attributes (hereinafter referred to as “experience article set DB”) is a DB having columns of experience article IDs, experience article texts, and experience attribute ID lists as shown in FIG.

The factor attribute DB 60 is a DB having factor attribute IDs and factor attribute labels as columns as shown in FIG.
The experience attribute DB 70 is a DB having experience attribute IDs and experience attribute labels as columns as shown in FIG.
Details of the processing of each part will be described below.

  FIG. 5 is a flowchart of the operation in one embodiment of the present invention.

  When there is an activation command (step 101), the experience article input unit 10 reads all article IDs and experience article texts of all articles from the experience article set DB 50 (step 102) and passes them to the word factor context score assignment unit 20. .

  The word factor context score assigning unit 20 receives all article IDs and all experience article texts from the attributed experience article set DB 50, and outputs all article IDs and word factor context scores of each word of all articles.

  As shown in FIG. 6, the word factor context score assigning unit 20 includes a document word feature calculating unit 21 and a factor context score assigning unit 22.

  The document word feature calculation unit 21 receives all article IDs and all experience article texts as input, and outputs all article IDs and scores of features of each word of all articles (step 103). The score of each feature of each word of an article is a score of a feature used when a factor context score is assigned to each subsequent word. As shown in FIG. 7, the document word feature calculation unit 21 gives a value of each feature to each word for each article. Features include the following: FIG. 8 is an example of clue expression in an embodiment of the present invention.

(i) an index representing the closeness of the appearance position of the word and a clue expression group (FIG. 8A) representing a factor context;
Here, examples of expressions representing experience (“articles written about going to hamburger shop A”) indicate “go”, “enter the store”, and the like.

(ii) an index indicating the proximity of the appearance position of the word and a clue expression group (FIG. 8B) indicating that it is not a factor context;
Here, examples of clues that are not written about the factors (“articles written about going to hamburger shop A”) refer to words that frequently appear after the decision to enter the store, such as “order” or “inside the store”.

(iii) an index indicating the position of the word in the document;
One or more indices are used to measure the closeness of the appearance positions between words.

The indicators for measuring the proximity of the above appearance positions include
(A) the reciprocal of the difference in the appearance position of phrases when the appearance phrases of each word are in a dependency relationship;
(B) the reciprocal of the difference in simple appearance position of each word;
(C) Reciprocal of difference in appearance position of each word (in the order of input, word A and word B) (however, when word A is before word B, it is 0);
(D) Reciprocal of difference in appearance position of each word (in the order of input, word A and word B) (however, when word B is before word A, it is 0);
and so on.

  The factor context score assigning unit 22 receives the total article ID and the score of each feature of each word of all articles, and outputs the total article ID and the word factor context score of each word of all articles (step 104). The word factor context score of a word in an article is an index indicating whether the word is used in a context representing a factor in the article. As a method of granting, a classifier is learned from positive examples given manually, and a method using machine learning that uses the certainty factor or a linear combination of the values of each feature of each word is used. And

  It supplements about the method using machine learning in said factor context score provision part 22. FIG. In the method using machine learning, an experience article set (each feature has been calculated) for training data is randomly created in advance. The class label “FACTOR” that indicates whether it is used in the context that represents the factor, and the class label “NOTFACTOR” that indicates whether it is used in the context that represents the factor, for each experience article word in the experience article set Is given manually. A classifier is created by learning in machine learning using the training data (note that a learning method that can calculate some certainty of which label is used). Using the classifier, a certainty factor is given to each unclassified data (each word of each document), and the certainty factor is used as a word factor context score.

  The experience article factor score assigning unit 30 reads the data of the factor attribute DB 60 and inputs all the article IDs from the word factor context score assigning unit 20 and the word factor context scores of all the words of all articles, as shown in FIG. For each article ID (1,..., Nd), a score is given for every factor attribute. For example, the score is given as a factor 0.3 of the factor ID 1 of the article ID 1, a score 0.2 of the factor ID 2,..., And the experience article factor score for each factor of all the article IDs and all the articles is given to the experience factor relation score calculating unit 40. Output (step 105). The factor attribute is determined in advance and is stored in the factor attribute DB 60 as in the example of FIG. The experience article factor score is learned in some way, such as by using a machine learning method that uses a certain classifier to learn a classifier from a manual example given by hand, or a rule that gives a relationship between a given appearance word and a factor attribute. Shall be given.

  The experience factor relationship score calculation unit 40 receives all the article IDs from the experience article factor score assigning unit 30 and the experience article factor score for each factor attribute of all articles, and outputs an all experience factor relationship score. The experience factor relationship score calculation unit 40 calculates an experience factor relationship score that is a value representing the strength of the relationship between the experience and the factors represented as shown in FIG.

  The experience factor relationship score calculation unit 40 calculates an experience factor relationship score by any of the following methods (step 106).

  As a first method, referring to the experience article set DB 50, for each experience attribute ID, as shown in FIG. Is calculated (step 106).

  As a second method, for each experience attribute ID, for each experience attribute ID, for each factor attribute, the sum of experience article factor scores of each experience article for each factor attribute (X1) and the corresponding experience The number of experience articles possessed as attributes (X2), the sum of experience article factor scores of the corresponding factor attribute IDs of all experience articles (X3), and the total number of articles read from the experience article set DB 50 (X4) Calculate each. At this time, for each experience attribute ID and factor attribute ID, an experience factor relationship between the experience attribute ID and the factor attribute ID by an index (P) that measures how much X1 for X2 is higher than X3 for X4 Output the score.

  The index P is a chi-square value in a 2 × 2 contingency table using X1, X2, X3, and X4 as shown in FIG. 12 (provided that 0 if X1 / X2 <X3 / X4) ) Can be used.

  Finally, the score calculation unit 42 refers to the factor attribute DB 60 and the experience attribute DB 70 as shown in FIG. 10, and displays each label and the experience factor relation score for each experience attribute ID and experience attribute ID as display means. Output (step 107).

  This experience factor-causal analysis device has an experience attribute of “hamburger” for experience article A (“Go to the neighborhood Mac !!” When you go to the shop, the strength of the relationship between the experience of going to the hamburger shop and the factor that you can use the power supply is output as 0.7.

  In the present invention, the operation of each component of the empirical factor causal analysis apparatus shown in FIG. 1 is constructed as a program, installed in a computer used as the empirical factor causal analysis apparatus, and executed. It is possible to circulate through.

  Further, according to the present invention, the constructed program can be stored in a portable storage medium such as a hard disk, a flexible disk, or a CD-ROM, and can be installed in a computer or distributed.

  The present invention is not limited to the above-described embodiments, and various modifications and applications are possible within the scope of the claims.

DESCRIPTION OF SYMBOLS 10 Experience article input part 20 Word factor context score provision part 21 Document word feature calculation part 22 Factor context score provision part 30 Experience article factor score assignment part 40 Experience factor relation score calculation part 50 Experience article set DB with attributes
60 Factor attribute DB
70 Experience attribute DB

Claims (9)

An experience factor relationship analyzer that outputs a score representing the strength of the relationship between various experiences and the various factors that are the main causes and conditions that people experience,
Experience article storage means with experience attributes storing experience articles that describe an experience article of people's own experience and an experience attribute that is an attribute indicating the type of experience of the experience article;
Experience attribute storage means storing experience attribute ID and experience attribute;
Factor attribute storage means storing factor attribute ID and factor attribute;
Experience article input means for reading an experience article set from the experience article storage means with experience attributes;
For each word of each experience article in the experience article set, a word factor context score giving means representing a factor of the degree that the word is used in the context representing the factor;
For each word of each experience article to which a word factor context score has been assigned, the factor attribute storage means is referenced to indicate what factor the experience described in each article by machine learning is based on An experience article factor score giving means for giving an experience article factor score;
Obtain an experience article set including the experience article factor score and experience attribute, obtain the experience article factor score, refer to the experience attribute storage means for each experience attribute of the experience article, and each experience and each factor Experience factor relationship output means for outputting a total experience factor relationship score indicating the strength of the relationship between
An empirical factor relationship analyzing apparatus characterized by comprising: The word factor context score assigning means includes:
When giving the word factor context score to each word of each experience article,
2. The empirical factor relationship analysis device according to claim 1, further comprising means for using the appearance position of each word in the document as a feature for assigning the word factor context score. The word factor context score assigning means includes:
When giving the word factor context score to each word of each experience article,
The means according to claim 1, further comprising means for using an appearance position relationship between each word and an experience expression (verb, noun, etc.) representing experience as a feature for assigning the word factor context score. Experience factor relationship analyzer. The word factor context score assigning means includes:
When giving the word factor context score to each word of each experience article,
2. The empirical factor relation analysis device according to claim 1, further comprising means for using a clue expression indicating that the factor is not written as a feature for assigning the word factor context score. The word factor context score assigning means includes:
When giving the word factor context score to each word of each experience article,
Means for using the appearance position of each word in the document as a feature for assigning the word factor context score;
Means for utilizing the position relationship between each word and an experience expression (with a verb or a noun) representing experience, as a feature that gives the word factor context score;
Means for using a clue expression indicating that the factor is not written as a feature for giving the word factor context score;
2. The empirical factor relationship analyzer according to claim 1, comprising at least one means. The experience article factor score giving means is:
For each experience attribute, which is attributed to each experience attribute, the sum of the experience article factor scores of each experience article is calculated for each factor, and the value of each sum of the experience attributes and factors is calculated. 2. The experience factor relationship analyzing apparatus according to claim 1, further comprising means for outputting the experience article factor score that is the strength of the relationship. The experience article factor score giving means is:
For each experience attribute that has it as an attribute, for each experience article set, the sum of experience article factor scores of each experience article for each factor (X1), the number of experience articles that have the experience as an attribute ( X2), the sum of the experience article factor scores (X3) and the total number of articles (X4) for the relevant factors of all experience articles, respectively. 2. The experience factor relationship analysis device according to claim 1, further comprising means for outputting a relationship between experience attributes and factors as the total experience factor relationship score by an index that measures how high the value is compared to X3 relative to X4. An experience factor relationship analysis method that outputs a score representing the strength of the relationship between various experiences and the various factors that are the main causes and conditions that cause people to experience,
Experience article storage means with experience attributes storing experience articles that describe an experience article of people's own experience and an experience attribute that is an attribute indicating the type of experience of the experience article;
Experience attribute storage means storing experience attribute ID and experience attribute;
In an apparatus having factor attribute ID and factor attribute storage means storing factor attributes,
Experience article input means reads an experience article set from experience article storage means with experience attribute storage means,
A word factor context score assigning step in which the word factor context score assigning unit assigns a word factor context score indicating the degree of use of each word of each experience article in the experience article set in the context representing the factor. When,
The experience article factor score assigning means refers to the factor attribute storage means for each word of each experience article to which the word factor context score has been assigned, and the experience described in each article is based on what factor An experience article factor score granting step for giving an experience article factor score representing whether or not
The experience factor relation output means acquires an experience article set including the experience article factor score and experience attribute, acquires the experience article factor score, and refers to the experience attribute storage means for each experience attribute of the experience article An experience factor relationship output step for outputting an overall experience factor relationship score indicating the strength of the relationship between each experience and each factor;
An experience factor relation analysis method characterized by   The program for functioning a computer as each means which comprises the experience factor relationship analysis apparatus of any one of Claims 1 thru | or 7.

Images