JP2010061176A - Text mining device, text mining method, and text mining program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To generate an attribute value or a combination of attribute values that is effective to a new text classification on the basis of a characteristic (of the text) selected by a user and that a user does not explicitly designate. <P>SOLUTION: In a text mining device, an attribute condition designating means 301 reads the conditions of an attribute value, i.e., conditions of the first class positive example or the first class negative example designated by the user via an input device 10. A text mining means 302 carries out text mining in the test of a test storage part 202, extracts effective portions to classify the first class positive example and the first class negative example as characteristics, stores the characteristics in a mining result holding part 203, and performs displaying via an output device 40. An analysis object characteristic designating means 303 inputs the characteristic selected by the user. A positive/negative example text extracting means 304 classifies the texts corresponding to the first class positive example and the first class negative example into the second class positive or negative text, based on the selected characteristic. An attribute characteristic extracting means 305 outputs the effective attribute value to classify the second class positive example and the second class negative example as characteristics. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a text mining device, a text mining method, and a text mining program for extracting a word as a text feature, and in particular, from a word obtained as a mining result, a characteristic attribute of the text including the word is extracted. The present invention relates to a text mining device, a text mining method, and a text mining program.

  Text mining appears when a user designates text with a specific attribute value as a positive example for a set of texts that have been given attribute values for several attributes. This is a process for extracting and outputting features to be output.

  For example, in the case of a response record at a contact center that accepts inquiries about products, in addition to text that describes the contents of accepted inquiries, “questions” and “requests” that indicate what kind of inquiries were received The type of inquiry such as “repair request”, the name of the model that was the subject of the inquiry, the date of receipt of the inquiry, and the name of the person in charge who responded to the inquiry are often recorded as a set. In text mining, such text is regarded as text with attribute values of “inquiry type” attribute, “model name” attribute, “acceptance date” attribute, and “person in charge” attribute. For example, “inquiry type” Using a text having “repair request” as an attribute value as a positive example, it is possible to extract features that appear biased in the text of the positive example.

  A conventional text mining device extracts words from each text, and extracts words or combinations of words that are highly related to text having specific attribute values specified as positive examples as features of the positive examples. Composed.

  An example of this type of text mining device is described in Patent Document 1. The text mining device described in Patent Document 1 sets a feature word extraction processing unit that extracts characteristic words and phrases appearing in a mining target text, and a classification axis (corresponding to an attribute) to be analyzed. An analysis axis setting processing unit, and a related phrase acquisition processing unit that extracts words / phrases having a high degree of association with each category (corresponding to attribute values) of the classification axis, and the classification axis set by the user as an analysis target Extract words that are characteristic of each category.

  Another example of this type of text mining method is described in Non-Patent Document 1. In the text mining method described in Non-Patent Document 1, when a positive example text (target group) and a negative example text (target group) are given, the appearance frequency in the positive example text is high, A pattern in the text that appears as low as possible in the text of the negative example, that is, a pattern that is effective in separating the positive example from the negative example is found by the appearance frequency, and the pattern is extracted as a feature of the positive example. Is.

  On the other hand, a technique for learning a certain pattern or rule from a data set other than text is called data mining, and various methods for performing data mining are widely known.

  As an example of a method for performing data mining, a branch conquest algorithm and a covering algorithm are described in Non-Patent Document 2. This method is a method for obtaining a decision tree for discriminating positive examples when there is a data set with attributes that are divided into positive examples and negative examples in advance.

  An example of another method for performing data mining is described in Non-Patent Document 3. This method is a method for obtaining a correlation rule between a set of items when a set of transactions that is a combination of items is given.

JP 2003-141134 A Junichiro Abe and four others, “High-speed data mining from text data: exploratory document browsing and application to web data”, Journal of the Japanese Society for Artificial Intelligence, Vol. 15, no. 4, July 2000, pp. 618-628 Hiroshi Motoda and two others, “Machine learning and data mining”, Journal of Artificial Intelligence, Vol. 12, no. 4, July 1997, pp. 505-512 Yu Kiteragawa, “Association rule extraction technique in data mining”, Journal of Artificial Intelligence, Vol. 12, no. 4, July 1997, pp. 513-520

  By text mining, if there is a feature that is noticeable in the text specified by the user as a positive example, it can be extracted. For this reason, when some feature is extracted, the user can know that the text specified as the positive example has a common feature. However, the extracted features do not always appear uniformly throughout the text designated as a positive example. Moreover, it does not necessarily appear only in the text designated as a positive example.

  For example, in response to a response record at a contact center that accepts inquiries about products, text mining is performed using the text with the date of acceptance “October 2005” and the type of inquiry “repair request” as a positive example. And the word “hard disk” is extracted. This means that the word “hard disk” appears in a biased manner in the text with the reception date “October 2006” and the inquiry type “repair request” compared to other texts. . From this result, the user can know that the repair requests in October 2005 were more related to “hard disks” than others.

  In this case, the word “hard disk” is actually a text with the model name “PC-100” among the texts with the reception date “October 2005” and the inquiry type “repair request”. There is a possibility that it appears in particular. In addition, the word “hard disk” may appear more frequently in the text with the reception date “November 2005” and the inquiry type “repair request” than others. However, the user has not been able to know it conventionally.

  As described above, the problem with the above-described conventional text mining apparatus is that when a feature is extracted from the text, the range of the text in which the feature appears cannot be presented to the user. That is, in the conventional text mining device, the attribute value (or combination of attribute values) effective for a new text classification not explicitly specified by the user based on the feature (text) selected by the user is known. Can not. The reason is that the above-described conventional text mining device does not present to the user information on what common features the text in which the extracted features appear has in addition to the appearance of the features. Because.

  An object of the present invention is to provide a text mining apparatus, a text mining method, and a text mining program that solve the above-described problems.

  The first text mining device of the present invention performs text mining based on an attribute value condition which is a condition of the first type positive example and the first type negative example specified by the user, and performs the first type positive example. A portion effective for classifying an example and a first type negative example is extracted as a feature, and a feature to be noticed is selected from the features, and the user selects the first type positive example, and the first type The text corresponding to the negative example is classified into the second type of positive text in which the selected feature appears and the second type of negative text in which the selected feature does not appear. It has a data processing device that generates an attribute value condition that is a new feature effective for classifying positive examples and second type negative examples.

  A second text mining device according to the present invention stores a plurality of texts and an attribute value for each text, reads the text and attribute values for each text from the storage device, and a user Text mining is performed by applying the attribute value condition, which is the condition of the first type positive example and the first type negative example, to the text and the attribute value for each text. A portion effective for classifying an example and a negative example of the first type is extracted as a feature, stored as a mining result in the storage device, and a feature to be noticed is selected from the features, and the user selects The text corresponding to one type of positive example and one type of negative example is the text of the second type of positive example in which the selected feature appears and the type of negative example in which the selected feature does not appear. Sort into text, It has two positive cases and the data processing device for outputting the generated output device a valid new features become attribute value condition to classify the second type of negative examples.

  The third text mining device of the present invention is the first or second text mining device, and has an effective part for classifying the first type positive example and the first type negative example. “A phrase having a high appearance frequency in the text of the first type positive example and a low appearance frequency in the text of the first type negative example” based on the first criterion set in advance.

  A fourth text mining device of the present invention is the first, second, or third text mining device for classifying a second type positive example and a second type negative example. An attribute value condition that becomes a valid new feature is based on a second criterion set in advance, “appears frequently as an attribute value for the second type positive example, and appears as an attribute value for the second type negative example. This is a combination of “attribute values with low frequency”.

  The fifth text mining device of the present invention inputs an attribute value condition that is a condition of the first type positive example and the first type negative example designated by the user, and the condition of the first type positive example Text mining is performed based on a certain attribute value condition, a portion effective for classifying the first type of positive example is extracted as a feature, the feature to be noted is selected from the features, and the user selects the first feature. The text corresponding to the positive example of the species and the negative example of the first type is classified into the text of the second type of positive example in which the selected feature appears and the text of the second type of negative example in which the selected feature does not appear. And a data processing device for generating an attribute value condition that is a new feature effective for classifying the second type positive example and the second type negative example.

  The sixth text mining apparatus of the present invention performs text mining based on an attribute value condition which is a first type of positive condition specified by the user, and converts the text matching the attribute value condition to the first type. As a positive example, a portion that is effective for classifying the remaining text as a first type negative example is extracted as a feature, the feature to be noticed is selected from the features, and the user selects the first type positive example. And the text corresponding to the first type negative example is classified into the second type positive example text in which the selected feature appears and the second type negative example text in which the selected feature does not appear. And a data processing device for generating an attribute value condition which is a new feature effective for classifying the second type positive example and the second type negative example.

  The seventh text mining device of the present invention extracts elements that appear frequently in all stored text as features, causes the user to select features to be noted from the features, and the selected features appear A data processing device that generates an attribute value condition that is a new feature effective for classifying positive examples and negative examples by separating them into positive example texts and negative example texts in which the selected feature does not appear Have.

  In the first text mining method of the present invention, the text mining device performs text mining based on an attribute value condition which is a condition of the first type positive example and the first type negative example designated by the user. A procedure for extracting, as a feature, a portion effective for classifying one positive example and a first negative example, a procedure for causing the user to select a feature to be noted from the extracted features, The text corresponding to one type of positive example and one type of negative example is the text of the second type of positive example in which the selected feature appears and the type of negative example in which the selected feature does not appear. And a procedure for generating an attribute value condition that is a new feature effective for classifying the second type positive example and the second type negative example.

  A second text mining method of the present invention is a text mining method in a text mining device comprising a plurality of texts, a storage device storing attribute values for each text, and a data processing device, wherein the data processing The apparatus reads out the text and the attribute value for each text from the storage device, and the attribute value condition which is the condition of the first type positive example and the first type negative example specified by the user. , Applying text mining to the text and the attribute value for each text, extracting as a feature a portion effective for classifying the first type positive example and the first type negative example, and storing the memory A procedure for storing the result as a mining result in the apparatus; a procedure for causing the user to select a feature to be noted from the extracted features; a first type positive example; Are classified into a second type of positive example text in which the selected feature appears and a second type of negative example text in which the selected feature does not appear. And a procedure for generating an attribute value condition as a new feature effective for classifying the example and the second type negative example and outputting the attribute value condition to the output device.

  A third text mining method of the present invention is the first or second text mining method, wherein a portion effective for classifying a first type positive example and a first type negative example is provided. “A phrase having a high appearance frequency in the text of the first type positive example and a low appearance frequency in the text of the first type negative example” based on the first criterion set in advance.

  A fourth text mining method of the present invention is the first, second, or third text mining method for classifying a second type positive example and a second type negative example. An attribute value condition that becomes a valid new feature is based on a second criterion set in advance, “appears frequently as an attribute value for the second type positive example, and appears as an attribute value for the second type negative example. This is a combination of “attribute values with low frequency”.

  According to a fifth text mining method of the present invention, the text mining device inputs an attribute value condition that is a condition of the first type positive example and the first type negative example designated by the user, Text mining based on the attribute value condition, which is a condition of the positive example of the seed, and a procedure for extracting a portion effective for classifying the positive example of the first kind as a feature, and a feature to be noticed from the feature The procedure selected by the user, the first type positive example, and the text corresponding to the first type negative example are selected as the second type positive example text in which the selected feature appears and the selected feature. To generate the attribute value condition which is a new feature effective for classifying the second type negative example and the second type negative example into the second type negative example text that does not appear Procedures.

  According to a sixth text mining method of the present invention, a text mining device performs text mining based on an attribute value condition which is a first type of positive condition designated by a user, and text conforming to the attribute value condition. Is a positive example of the first type, and a procedure for extracting as a feature a portion effective for classifying the remaining text as a negative example of the first type, and causing the user to select a feature to be noted from among the features The procedure, the first type positive example, and the text corresponding to the first type negative example, the second type positive example text in which the selected feature appears and the second type in which the selected feature does not appear And a procedure for generating an attribute value condition as a new feature effective for classifying the second type positive example and the second type negative example.

  According to a seventh text mining method of the present invention, the text mining device causes the user to select a feature to be noticed from among a procedure in which elements frequently appearing in all stored texts are extracted as features. Separating the procedure into positive example text in which the selected feature appears and negative example text in which the selected feature does not appear, and an attribute that becomes a new feature effective for classifying the positive example and the negative example Generating a value condition.

  The first text mining program of the present invention performs text mining based on an attribute value condition which is a condition of the first type positive example and the first type negative example specified by the user, and the first type positive example. And a procedure for extracting a portion effective for classifying the first type negative example as a feature, a procedure for causing the user to select a feature to be noted from the extracted features, and a first type positive example , And the text corresponding to the first type negative example is classified into the second type positive example text in which the selected feature appears and the second type negative example text in which the selected feature does not appear. The text mining apparatus is caused to execute a procedure and a procedure for generating an attribute value condition that is a new feature effective for classifying the second type positive example and the second type negative example.

  A second text mining program of the present invention is a text mining program in a text mining device comprising a plurality of texts, a storage device storing attribute values for each text, and a data processing device, wherein the storage device A procedure for reading the text and the attribute value for each text, and an attribute value condition which is a condition of the first type positive example and the first type negative example designated by the user, the text, and Text mining is applied to the attribute value for each text, and a portion effective for classifying the first type positive example and the first type negative example is extracted as a feature and stored as a mining result in the storage device A procedure for selecting the feature to be noticed from the extracted features, a first type positive example, and a first type negative example The corresponding text is classified into a second type of positive example text in which the selected feature appears and a second type of negative example text in which the selected feature does not appear. The data processing apparatus is caused to execute a procedure for generating an attribute value condition as a new feature effective for classifying the negative example of the seed and outputting the attribute value condition to the output apparatus.

  The third text mining program of the present invention is the first or second text mining program, wherein the effective part for classifying the first type positive examples and the first type negative examples is provided. “A phrase having a high appearance frequency in the text of the first type positive example and a low appearance frequency in the text of the first type negative example” based on the first criterion set in advance.

  A fourth text mining program according to the present invention is the first, second, or third text mining program for classifying the second type positive example and the second type negative example. An attribute value condition that becomes a valid new feature is based on a second criterion set in advance, “appears frequently as an attribute value for the second type positive example, and appears as an attribute value for the second type negative example. This is a combination of “attribute values with low frequency”.

  The fifth text mining program of the present invention includes a procedure for inputting an attribute value condition which is a condition of a first type positive example and a first type negative example designated by the user, and a first type positive example. Text mining based on the attribute value condition that is a condition, a procedure for extracting a portion effective for classifying the first type of positive example as a feature, and selecting a feature to be noticed from the features to the user A second type of positive example in which the selected feature appears and a second feature in which the selected feature does not appear, and the procedure corresponding to the first type positive example and the text corresponding to the first type negative example Text mining includes a procedure for separating text into negative example text and a procedure for generating an attribute value condition that is a new feature effective for classifying the second type positive example and the second type negative example. Let the device run.

  The sixth text mining program of the present invention performs text mining on the basis of an attribute value condition which is a first type of positive condition specified by the user, and converts the text matching the attribute value condition to the first type. A procedure for extracting, as a feature, a portion effective for classifying the remaining text as a first type negative example, a procedure for causing the user to select a feature to be noted from the features, The text corresponding to the positive example of the species and the negative example of the first type is classified into the text of the second type of positive example in which the selected feature appears and the text of the second type of negative example in which the selected feature does not appear. And a procedure for generating an attribute value condition, which is a new feature effective for classifying the second type positive example and the second type negative example, is executed by the text mining apparatus.

  The seventh text mining program of the present invention includes a procedure for extracting, as a feature, an element that frequently appears in all stored text, a procedure for causing the user to select a feature to be noted from the features, The attribute value condition that becomes a new feature effective for classifying the positive example and the negative example is generated by separating the positive example text in which the feature appears and the negative example text in which the selected feature does not appear. The procedure is executed by a text mining device.

  An eighth text mining device of the present invention is a text mining device that extracts and outputs features from a set of texts with attributes, and includes analysis target feature designating means for inputting features to be noted from the features. A positive example negative example text extracting means for extracting a positive example text and a negative example text from the text depending on whether or not the inputted feature appears in the text, and the positive example text and the negative example text, Attribute feature extraction means for extracting attribute features effective for classification.

  According to a ninth text mining device of the present invention, a text storage means for holding a set of text, an attribute storage means for holding an attribute value for the text, a condition designating means for inputting a text mining condition, and according to the condition Text mining means for extracting text features, analysis target feature designating means for inputting features to be noticed from among the features, and whether or not the inputted features appear in the text, positive examples from the text Positive example negative example text extracting means for extracting text and negative example text, and attribute feature extracting means for extracting attribute features effective for classifying the positive example text and the negative example text .

  An eighth text mining method of the present invention is a text mining method in which a computer extracts and outputs features from a set of texts with attributes, and the computer inputs features to be noted from the features. The computer extracts positive text and negative text from the text according to whether the input feature appears in the text, and classifies the positive text and the negative text The computer extracting attribute features that are useful for the computer.

  According to a ninth text mining method of the present invention, a computer stores a set of texts and attribute values for the text, the text mining conditions are input to the computer, and text features are extracted according to the conditions. A step of inputting a feature to be noticed from among the features, and a step of extracting positive example text and negative example text from the text depending on whether or not the inputted feature appears in the text; And extracting an attribute characteristic effective for classifying the positive example text and the negative example text.

  An eighth text mining program of the present invention is a text mining program for causing a computer to execute processing for extracting and outputting features from a set of texts with attributes, and inputting features to be noted from the features. Analysis target feature designation processing, positive example negative example text extraction processing for extracting positive example text and negative example text from the text, depending on whether the input feature appears in the text, the positive example text, The computer is caused to execute attribute feature extraction processing for extracting attribute features effective for classifying the negative example text.

  According to a ninth text mining program of the present invention, a process for storing a set of texts and an attribute value for the text in a storage device, a condition designating process for inputting text mining conditions, and extracting text features according to the conditions Depending on whether the input feature appears in the text, the positive text and the negative text from the text And a positive example negative example text extraction process, and an attribute feature extraction process that extracts an attribute characteristic effective for classifying the positive example text and the negative example text.

  The effect of the present invention is that the user knows attribute values (or combinations of attribute values) effective for a new text classification that is not explicitly specified by the user based on the feature (text) selected by the user. It can be done.

  The reason for this is that, among text features extracted by text mining based on text attribute values specified by the user, the text that the user selected appears as a positive example, and the text that does not appear as a negative example This is because mining is performed and attribute values or combinations of attribute values effective for classifying positive examples and negative examples are extracted and output.

  First, an outline of the present invention will be described. The text mining device of the present invention performs text mining based on an attribute value condition which is a condition of the first type positive example and the first type negative example specified by the user, and the first type positive example and the first type positive example. A portion effective for classifying one kind of negative example is extracted as a feature, and a feature to be noted is selected from the features.

  Next, the text mining device determines that the text corresponding to the first type positive example and the first type negative example are the text of the second type positive example in which the selected feature appears and the selected feature is selected. The attribute value condition which becomes a new feature effective for classifying the second type positive example and the second type negative example is generated by classifying the text into the second type negative example text that does not appear.

  Here, the “effective portion for classifying positive examples and negative examples” is, for example, “a phrase having a high appearance frequency in the positive example text and a low appearance frequency in the negative example text”. . That is, the phrase is not limited to “a phrase that appears in positive text and does not appear in negative text”. Further, for example, whether the appearance frequency is high or the appearance frequency is low can be determined by comparison with respective “threshold values” set in advance. Further, for example, it is possible to determine from the ratio of the frequency of appearing in the positive example text to the frequency of appearing in the negative example text. As described above, the appearance frequency may be determined based on a predetermined criterion. The classification can be based on various scales other than the appearance frequency. Hereinafter, “classification” is used in the above meaning.

  Next, a first embodiment of the present invention will be described in detail with reference to the drawings. FIG. 1 is a block diagram showing the configuration of the first exemplary embodiment of the present invention. Referring to FIG. 1, a text mining device according to a first embodiment of the present invention includes an input device 10 such as a keyboard and a mouse, a storage device 21 such as a hard disk for storing information, and a data processing device that operates under program control. 31 and an output device 40 such as a display device.

  The storage device 21 includes an attribute storage unit 201, a text storage unit 202, and a mining result holding unit 203. The attribute storage unit 201 stores information on attribute values assigned to the text in association with each text stored in the text storage unit 202. The text storage unit 202 stores text that is a target of text mining.

  FIG. 2 shows an example of the text storage unit 202, and FIG. 3 shows an example of the attribute storage unit 201. In this example, a unique text number is assigned to each text and stored in the text storage unit 202. In the attribute storage unit 201, “inquiry type”, “model name”, “reception date”, “in charge” for each text number. The attribute values of the four types of attributes are stored.

  Note that the attribute storage unit 201 and the text storage unit 202 need not be completely separated, and may be configured to store the text and the attribute for the text at the same time. The mining result holding unit 203 stores features obtained as a result of text mining performed on the text stored in the text storage unit 202.

  The data processing device 31 includes attribute value condition specifying means 301, text mining means 302, analysis target feature specifying means 303, positive / negative example text extracting means 304, and attribute feature extracting means 305. The attribute value condition designating unit 301 reads the positive value (first type) attribute value condition and the negative example (first type) attribute value condition specified by the user through the input device 10. .

  The text mining means 302 converts the text stored in the text storage unit 202 into a positive example text and a negative example attribute value that match the positive example attribute value condition read by the attribute value condition specifying unit 301. Text mining is applied as negative example text that meets the conditions. As a result, the text mining means 302 extracts a feature effective for classifying the positive example as the negative example as the feature of the positive example text, and outputs it to the user through the output device 40. Further, the extracted feature is stored in the mining result holding unit 203.

  In text mining, in general, elements constituting a part of text such as a word, a set of a plurality of words, a phrase, a sentence, and the like are extracted as features. That is, in the text mining, for example, those elements that do not appear so much in the negative example text and appear biased in the positive example text are extracted as features of the positive example text. The technology described in Non-Patent Document 1 can be partially applied to this text mining.

  There is a text mining technique in which the structure of text is analyzed, the text is converted into structured data as a result of the analysis, and then the partial structure of the structured data is extracted as a feature. This is because, for example, the dependency relationship between words is analyzed in advance, and two words in the dependency relationship are extracted as features, or the text is converted into a dependency structure tree by dependency structure analysis. For example, a method of extracting a subtree as a feature. When such a method is used, when a partial structure is included in structured data obtained from text, it is considered that the partial structure appears in the text.

  The text mining means 302 outputs the characteristics obtained by text mining to the user through the output device 40 and stores them in the mining result holding unit 203. The information output to the user through the output device 40 includes the number of texts in which the features appear in addition to the extracted features, how much the features are biased in the text of the normal example, etc. Additional information may be included.

  The analysis target feature designating unit 303 causes the user to designate a feature to be noted among the features output by the text mining unit 302 and reads the designated content through the input device 10.

  The positive example negative example text extraction unit 304 is a text to be processed by the text mining unit 302 among the texts stored in the text storage unit 202, that is, a positive example read by the attribute value condition designating unit 301 or For each of the texts that meet any of the negative example attribute value conditions, it is determined whether or not the feature read by the analysis target feature designating unit 303 appears, and the text in which the feature appears is determined to be a positive example (the above-described second example). The text that does not have the feature (of the seed) is extracted as a negative example (the above-mentioned second kind).

  In order to speed up the discrimination between the positive example and the negative example by the positive example / negative example text extraction unit 304, the text mining unit 302 creates and records an index indicating in which text each feature appears. Alternatively, the positive example negative example text extraction means 304 may determine the positive example and the negative example with reference to the index.

  When the analysis target feature designating unit 303 causes the user to designate a feature, only one feature or a plurality of features may be designated. When the analysis target feature designating unit 303 causes the user to designate a plurality of features, the positive example negative example text extracting unit 304 may use the text in which any one of the features appears as a positive example, A text in which a feature appears may be a positive example.

  In addition, when the positive example negative example text extraction unit 304 discriminates between the positive example and the negative example, only the text that appears the number of times the feature read by the analysis target feature designating unit 303 exceeds a certain threshold is discriminated as a positive example. May be.

  The attribute feature extraction unit 305 applies data mining to the positive example text and negative example text extracted by the positive example negative example text extraction unit 304 and classifies the positive example text and the negative example text. A characteristic attribute value or a combination of attribute values that is effective for the above is extracted and output to the user through the output device 40.

  In the first embodiment of the present invention, the data mining technique applied by the attribute feature extraction unit 305 is not limited to a specific method.

  For example, a decision tree analysis technique can be used as a data mining technique for extracting attribute values or combinations of attribute values that are characteristic of positive text. That is, a decision tree having a branch condition that is a combination of attribute values for classifying positive example text and negative example text is obtained, and the combination of attribute values when following the path leading to the positive example in the decision tree is obtained. It can be extracted as a combination of text-specific attribute values. In addition, a decision tree can be calculated | required using the method described in the nonpatent literature 2, for example.

  Similarly, as a data mining technique for extracting attribute values or combinations of attribute values that are characteristic of positive text, for example, a correlation analysis technique can be used. That is, when the set of positive texts is expressed as Tp, the set of texts having the condition V based on the combination of attribute values is expressed as T (V), and the number of texts belonging to the text set X is expressed as N (X), the confidence C ( V) = N (Tp∩T (V)) / N (T (V)) is higher than a predetermined threshold Cth, and the support degree S (V) = N (Tp∩T (V)) is a predetermined threshold. When it is higher than Sth, the combination of attribute values represented by V is extracted as a combination of attribute values specific to the text of the positive example. Since this corresponds to extracting an association rule that satisfies the minimum support level and the minimum certainty level, it can be realized by, for example, the method described in Non-Patent Document 3.

  In addition, any data mining technique can be used as long as it is a technique that can extract attribute values or combinations of attribute values that are characteristic of positive text.

  Next, the operation of the first embodiment of the present invention will be described in detail with reference to FIG. 1 and FIG. FIG. 4 is a flowchart showing the operation of the first exemplary embodiment of the present invention.

  First, the attribute value condition designating unit 301 reads the attribute value condition designated by the user as the positive and negative example conditions via the input device 10 (step A1 in FIG. 4).

  Next, the text mining unit 302 determines that the text stored in the text storage unit 202 matches the positive example attribute value condition read by the attribute value condition specifying unit 301 as the positive example text and the positive example. Text mining is performed with the text that satisfies the attribute value condition as negative example text, and features effective for classifying the positive example text and the negative example text are extracted (step A2).

  The text mining means 302 stores the extracted features in the mining result holding unit 203, reads the extracted features from the mining result holding unit 203, and outputs them to the user through the output device 40 (step A3). Next, the analysis target feature designating unit 303 reads the feature selection by the user via the input device 10 (step A4).

  The positive example negative example text extraction unit 304 reads the text stored in the text storage unit 202 one by one (step A5), and the text is a positive example or negative example attribute read by the attribute value condition specifying unit 301. It is determined whether or not any of the value conditions is met (step A6). If they match (step A6 / Yes), the positive / negative example text extraction unit 304 determines whether or not the feature selected by the user in step A4 appears in the text (step A7). When a feature appears in the read text (step A7 / Yes), the positive example negative example text extraction unit 304 sets the text as a positive example (step A8), and when no feature appears (step A7 / No), the text is a negative example (step A9). The positive example negative example text extraction unit 304 repeats the processing of Steps A5-A9 until all the texts have been processed (Step A10).

  Next, the attribute feature extraction unit 305 extracts, by data mining, an attribute value or a combination of attribute values effective for classifying the positive example text and the negative example text extracted by the processing in steps A5-A10. (Step A11). Next, the attribute feature extraction means 305 outputs the extraction result (attribute value or combination of attribute values) to the user via the output device 40 (step A12).

  In the first embodiment of the present invention, the attribute value condition specifying unit 301 reads the positive example attribute value condition and the negative example attribute value condition specified by the user, and the text mining unit 302 sets the positive value attribute value condition. Text mining is performed with text that meets the attribute value condition of the example as a positive example and text that meets the attribute value condition of a negative example as a negative example. Unlike this, the attribute value condition designating unit 301 receives only positive example attribute value conditions from the user, and the text mining unit 302 treats all texts that do not meet the positive example attribute value conditions as negative example texts. Is also possible. In this case, the positive example negative example text extraction unit 304 extracts the positive example text and the negative example text for all the texts stored in the text storage unit 202.

  In addition, the attribute mining condition specifying unit 301 is not provided, and the text mining unit 302 includes elements that frequently appear in all the texts stored in the text storage unit 202 (words, sets of words, phrases, sentences, etc.). A configuration to extract is possible. Also in this case, the positive example negative example text extraction unit 304 extracts the positive example text and the negative example text for all the texts stored in the text storage unit 202.

  Next, effects of the first exemplary embodiment of the present invention will be described.

  In the first embodiment of the present invention, among the features of the text extracted by text mining based on the text attribute values that are the conditions of the positive example (first type) and the negative example (first type), the user Data mining is performed with the text in which the selection of appears appears as a positive example (second type) and the text that does not appear as a negative example (second type), and a positive example (second type) and a negative example (second type) Attribute values or combinations of attribute values that are effective for classifying and outputting.

  In other words, in the first embodiment of the present invention, the attribute specific to the text in which the feature (text) selected by the user appears (not necessarily limited to the text in which all the selected features appear). Unique features to the user.

  Therefore, according to the first embodiment of the present invention, the user can create a new text classification (second type positive example, second type) that is not explicitly specified by the user based on the feature (text) selected by the user. It is possible to know attribute values (or combinations of attribute values) effective for two types of negative examples).

  Next, a second embodiment of the present invention will be described in detail. The configuration of the second embodiment of the present invention in FIG. 1 is the same as the configuration of the first embodiment of the present invention. The second embodiment of the present invention is a condition of the first type positive example among the attribute value conditions that are conditions of the first type positive example and the first type negative example designated by the user. Text mining is performed based on the attribute value condition, a portion effective for classifying the first type positive example from the entire text is extracted as a feature, and a feature to be noticed is selected from the features.

  Next, the text mining device determines that the text corresponding to the first type positive example and the first type negative example are the text of the second type positive example in which the selected feature appears and the selected feature is selected. The attribute value condition which becomes a new feature effective for classifying the second type positive example and the second type negative example is generated by classifying the text into the second type negative example text that does not appear.

  The second embodiment of the present invention has a simpler configuration than the first embodiment of the present invention because the text mining means 302 only needs to perform mining based on the first positive example. Has the effect.

  Next, a third embodiment of the present invention will be described in detail with reference to the drawings. FIG. 5 is a block diagram showing the configuration of the third exemplary embodiment of the present invention. Referring to FIG. 5, the third embodiment of the present invention includes an input device 10, a storage device 22, a data processing device 32 (for example, a computer), an output device 40, and a text mining program 50.

  The text mining program 50 includes an attribute value condition specifying unit 301, a text mining unit 302, an analysis target feature specifying unit 303, a positive / negative example text extracting unit 304, and an attribute according to the first and second embodiments of the present invention. The function of the feature extraction unit 305 is realized. The text mining program 50 is stored in the storage device 22 or other storage means (not shown).

  The text mining program 50 is read and executed by the stored data processing device 32 and controls the operation of the data processing device 32. The data processing device 32 executes the same processing as that of the data processing device 31 in the first and second embodiments under the control of the text mining program 50.

  As described above, the third embodiment of the present invention has an effect that it is easy to implement because the processing of FIG. 4 is executed by cooperation of hardware and software.

  Next, examples of the first embodiment of the present invention will be described in detail with reference to the drawings. Here, the operation of the embodiment of the present invention will be described by taking as an example the case where the characteristics of the inquiry in October 2005 are extracted for the inquiry about the repair request from the response records in the contact center that accepts the inquiry about the product. To do.

  As shown in FIG. 3, the attribute storage unit 201 stores attribute values of four types of attributes of “inquiry type”, “model name”, “acceptance date”, and “person in charge” for each text. As shown in FIG. 2, the text storage unit 202 stores in advance text to be text mined (contents of reception records).

  At this time, first, the attribute value condition designating unit 301 reads the designation of the attribute value condition of the text mining positive example and negative example by the user through the input device 10.

  Here, the user sets “(inquiry type” = “repair request”) AND (“reception date” = “October 2005”) ”as a positive example (first type) attribute value condition, and a negative example The condition of “(“ inquiry type ”=“ repair request ”) AND (“ reception date ”≠“ October 2005 ”)” is designated as the (first type) attribute value condition.

  Next, the text mining means 302 uses the text stored in the text storage unit 202 with “inquiry type is repair request” and “reception date is October 2005” as a positive example. Effective for classifying positive example text and negative example text by executing text mining with “inquiry type is repair request” and “reception date is not October 2005” as a negative example Extract features.

  In the text of T1-T7 recorded in the text storage unit 202 in FIG. 2, T1, T5, and T7 are positive examples (first type), and T6 is a negative example (first type). The text of T2-T4 is not used for text mining because neither the positive attribute value condition nor the negative attribute value condition is satisfied. The text mining means 302 outputs the extracted feature to the user via the output device 40 and stores it in the mining result holding unit 203.

  FIG. 6 is an explanatory diagram illustrating an example of the result of text mining. Here, the text mining means 302 extracts words appearing in the text as features, and stores the features as shown in FIG. 6 in the mining result holding unit 203. Next, the analysis target feature designating unit 303 causes the user to select a feature of interest and reads the selected content via the input device 10.

  For example, the analysis target feature designating unit 303 can input whether or not to select each feature output by the text mining unit 302 and can cause the user to select the feature.

  FIG. 7 is an explanatory diagram illustrating an example of contents displayed on the output device 40. Referring to FIG. 7, the analysis target feature designating unit 303 displays a check box for indicating that the user has selected the feature for each feature extracted by the text mining unit 302, and the user Read the feature with the check box checked. In FIG. 7, the word “hard disk” and the word “HDD” are selected by the user.

  The positive example negative example text extraction unit 304 is a positive example (first type) or negative example (first type) read by the attribute value condition specifying unit 301 among the texts stored in the text storage unit 202. For each text that meets one of the attribute value conditions, it is determined whether or not the feature specified by the user appears. If the feature appears, the feature does not appear as a positive example (type 2). In this case, it is extracted as a negative example (second type).

  Here, when any of the features specified by the user appears, the text is extracted as a positive example.

  According to FIG. 2, since the text of T1 matches the attribute value condition of the positive example (first type) read by the attribute value condition specifying unit 301 and includes the word “hard disk”, the positive example (second) Extracted as seeds). On the other hand, the text of T2-T4 does not correspond to the attribute value condition of the positive example (first type) nor the attribute value condition of the negative example (second type). Is not extracted as a negative example (second type).

  The text of T5 conforms to the attribute value condition of the positive example (first type) read by the attribute value condition specifying unit 301, but does not include the word “hard disk” nor the word “HDD”. Extracted as seeds).

  The text of T6 is extracted as a positive example (second type) because it matches the negative example (first type) attribute value condition read by the attribute value condition designating unit 301 and includes the word “HDD”. Is done. The text of T7 is extracted as a positive example (second type) because it matches the attribute value condition of the positive example (first type) read by the attribute value condition specifying unit 301 and includes the word “HDD”. Is done. The same processing is performed for other texts.

  Next, the attribute feature extraction unit 305 performs data mining on the positive example (second type) and negative example (second type) text extracted by the positive example negative example text extraction unit 304. Applying and extracting an effective attribute value or combination of attribute values to classify positive example (second type) text and negative example (second type) text, and output the output result to the output device 40 To the user.

  In the present embodiment, a decision tree that classifies positive example (second type) text and negative example (second type) text using a combination of attribute values as a branching condition is obtained and determined by data mining. A combination of attribute values corresponding to the path leading to the positive example (second type) in the tree is extracted as a combination of attribute values that are characteristic of the text of the positive example (second type). FIG. 8 is an explanatory diagram illustrating an example of a decision tree.

  Assuming that a decision tree as shown in FIG. 8 is obtained by data mining, a combination of attribute values that are characteristically seen in the text of the positive example (second type) is “(“ reception date ”=“ 2005 ”. “October” OR “November 2005”) AND (“model name” = “PC-100”) ”is obtained.

  FIG. 9 is an explanatory diagram showing an output example of the attribute feature extraction unit 305 in this case. Referring to FIG. 9, the combination of the output attribute values is “(“ reception date ”=“ October 2005 ”OR“ November 2005 ”) AND (“ model name ”=“ PC-100 ”). ].

  FIG. 10 is an explanatory diagram showing the logic of this embodiment. Referring to FIG. 10, in this embodiment, the user sets, as a positive example (first type), a text whose inquiry type is a repair request text with a reception date of October 2005 (FIG. 10 (a)). ) R11), and non-October 2005 as a negative example (first type) (R10 in FIG. 10 (a)), text mining, and "Hard disk" as a feature of the October 2005 repair request , “OS”, “HDD”, “error”, etc.

  Next, among these characteristics, data mining is performed by focusing on the “hard disk” and “HDD” selected by the user. As an attribute characteristic of a positive example (second type) text (R21 in FIG. 10B) in which “hard disk” or “HDD” appears, “(“ reception date ”=“ October 2005 ”OR“ 2005 ” “November, year”) AND (“model name” = “PC-100”) ”combination of attribute values is obtained.

  As a result, the condition initially specified by the user as a positive example was that the inquiry type is a repair request and the reception date is October 2005. The word “HDD” appears not only in October 2005 but also in the text of the repair request in November 2005, and is particularly prominent in the repair request text, especially in the PC-100 model. You can know that it has appeared.

  According to the present invention, a mining system for extracting effective knowledge such as defect information and problems from inquiry data recorded at a call center and paper document data such as a report, and a program such as a program for realizing the mining system Applicable to. Further, the present invention is applicable to a system or the like that accumulates the contents of inquiries as text and extracts frequently inquiries from the contents to construct a Q & A collection.

The block diagram which shows the structure of the 1st Embodiment of this invention. Explanatory drawing which shows the example of the content of a text memory | storage part. Explanatory drawing which shows the example of the content of an attribute memory | storage part. The flowchart which shows the operation | movement of the 1st Embodiment of this invention. The block diagram which shows the structure of the 2nd Embodiment of this invention. Explanatory drawing which shows an example of the result of text mining. Explanatory drawing which shows an example of the content displayed on an output device. Explanatory drawing which shows an example of a decision tree. Explanatory drawing which shows an example of the output of an attribute characteristic extraction means. Explanatory drawing which shows the logic of the Example of the 1st Embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Input device 40 Output device 50 Text mining program 21 Storage device 22 Storage device 31 Data processing device 32 Data processing device 201 Attribute storage unit 202 Text storage unit 203 Mining result holding unit 301 Attribute value condition specifying unit 302 Text mining unit 303 Analysis object Feature designation means 304 Positive example negative example text extraction means 305 Attribute feature extraction means

Claims (27)

Text mining is performed based on the attribute value condition which is the condition of the first type positive example and the first type negative example specified by the user, and the first type positive example and the first type negative example are classified. The effective part is extracted as a feature, the feature to be noticed is selected from the features, and the user selects the feature corresponding to the first type positive example and the first type negative example. The second type of positive example text in which the selected feature appears and the second type of negative example text in which the selected feature does not appear are classified into the second type positive example and the second type negative example. A text mining device comprising a data processing device for generating an attribute value condition that is a new feature effective for classification. A storage device that stores a plurality of texts and attribute values for each of the texts, reads the texts and attribute values for each of the texts from the storage device, and is a first type of positive example designated by the user, Applying the attribute value condition, which is the condition of the first type negative example, to the text and the attribute value for each text, and performing text mining to classify the first type positive example and the first type negative example A portion that is effective to do this is extracted as a feature, stored as a mining result in the storage device, and the user is allowed to select a feature to be noticed from the features, and the first type positive example and the first type Are classified into a second type of positive example text in which the selected feature appears and a second type of negative example text in which the selected feature does not appear. To classify examples and negative type 2 Text mining apparatus, comprising a data processing apparatus for outputting the generated attribute value condition that is valid new feature output device. The portion that is effective for classifying the first type positive example and the first type negative example is based on the first criterion set in advance, “the appearance frequency of the first type positive example is high. The text mining device according to claim 1, wherein the phrase is a phrase having a low appearance frequency in the first type negative example text. An attribute value condition that is a new feature effective for classifying the second type positive example and the second type negative example is based on a second criterion set in advance. 4. The text mining device according to claim 1, 2, or 3, wherein the attribute value is a combination of an attribute value having a high appearance frequency as an attribute value and a low appearance frequency as an attribute value for the second type negative example. Input the attribute value condition that is the condition of the first type positive example and the first type negative example specified by the user, and perform text mining based on the attribute value condition that is the condition of the first type positive example , Extracting a portion effective for classifying the first type positive example as a feature, allowing the user to select a feature to be noted from among the features, the first type positive example, and the first type The text corresponding to the negative example is classified into the text of the second type positive example in which the selected feature appears and the text of the second type negative example in which the selected feature does not appear, and the second type positive example And a data processing device for generating an attribute value condition which is a new feature effective for classifying the second type negative example. Text mining is performed based on the attribute value condition that is the condition of the first type positive example specified by the user, the text that matches the attribute value condition is set as the first type positive example, and the remaining text is set as the first type. The part effective to classify as a negative example is extracted as a feature, the feature to be noticed is selected from the features, and the user is selected, and corresponds to the first type positive example and the first type negative example The second type positive example text in which the selected feature appears and the second type negative example text in which the selected feature does not appear are classified into the second type positive example and the second type text. A text mining device comprising a data processing device for generating an attribute value condition which is a new feature effective for classifying negative examples of Elements that frequently appear in all stored texts are extracted as features, and the user is allowed to select a feature to be noticed from among the features, and a positive example text in which the selected feature appears and the selected feature are A text mining device comprising: a data processing device that generates an attribute value condition which is a new feature effective for classifying positive examples and negative examples by classifying them into negative example texts that do not appear. The text mining device performs text mining based on an attribute value condition which is a condition of the first type positive example and the first type negative example specified by the user, and the first type positive example and the first type negative example. A procedure for extracting a portion effective for classifying an example as a feature, a procedure for causing the user to select a feature to be noted from the extracted features, a first type positive example, and a first type Separating the text corresponding to the negative example into the second type positive example text in which the selected feature appears and the second type negative example text in which the selected feature does not appear, and the second type And a procedure for generating an attribute value condition that is a new feature effective for classifying the positive example and the negative example of the second type. A text mining method in a text mining device comprising a plurality of texts, a storage device for storing attribute values for each text, and a data processing device,
The data processing device is a procedure for reading the text and the attribute value for each text from the storage device, and attributes that are conditions of the first type positive example and the first type negative example specified by the user Text mining is performed by applying the value condition to the text and the attribute value for each text, and a portion effective for classifying the first type positive example and the first type negative example is extracted as a feature. Corresponds to the procedure for storing the result as the mining result in the storage device, the procedure for allowing the user to select the feature to be noticed from the extracted features, the first type positive example, and the first type negative example The second type positive example text in which the selected feature appears and the second type negative example text in which the selected feature does not appear are classified into the second type positive example and the second type text. New features useful for classifying negative cases Text mining method which comprises the steps of outputting the generated output device attribute value condition that. The portion that is effective for classifying the first type positive example and the first type negative example is based on the first criterion set in advance, “the appearance frequency of the first type positive example is high. 10. The text mining method according to claim 8, wherein the phrase is a phrase having a low appearance frequency in the first type negative example text. An attribute value condition that is a new feature effective for classifying the second type positive example and the second type negative example is based on a second criterion set in advance. The text mining method according to claim 8, wherein the attribute value is a combination of “attribute values having high appearance frequency as attribute values and low appearance frequency as attribute values for the second type negative example”. A procedure in which the text mining device inputs an attribute value condition that is a condition of the first type positive example and a first type negative example designated by the user, and an attribute value condition that is a condition of the first type positive example A procedure for performing text mining based on the above and extracting a portion effective for classifying the first type of positive example as a feature, a procedure for causing the user to select a feature to be noted from the features, The text corresponding to the positive example of the species and the negative example of the first type is classified into the text of the second type of positive example in which the selected feature appears and the text of the second type of negative example in which the selected feature does not appear. And a procedure for generating an attribute value condition that is a new feature effective for classifying the second type positive example and the second type negative example. Mining method. The text mining device performs text mining based on the attribute value condition that is the condition of the first type of positive example specified by the user, and sets the text that matches the attribute value condition as the first type of positive example, and the remaining A procedure for extracting, as a feature, a portion effective for classifying a text as a first type negative example, a procedure for causing the user to select a feature to be noted from among features, a first type positive example, and Separating the text corresponding to the first type negative example into the second type positive example text in which the selected feature appears and the second type negative example text in which the selected feature does not appear; And a procedure for generating an attribute value condition that is a new feature effective for classifying the second type positive example and the second type negative example. A procedure in which the text mining device extracts frequently appearing elements in all stored texts as features, a procedure in which the user selects a feature to be noted from among the features, and a correct feature in which the selected feature appears. Separating the example text from the negative example text in which the selected feature does not appear, and generating a new feature attribute value condition that is useful for classifying the positive and negative examples. A featured text mining method. Text mining is performed based on an attribute value condition which is a condition of the first type positive example and the first type negative example specified by the user, and the first type positive example and the first type negative example are classified. Applicable to the procedure for extracting the effective part of the feature as the feature, the procedure for allowing the user to select the feature to be noticed from the extracted feature, the first type positive example, and the first type negative example The second type positive example text in which the selected feature appears and the second type negative example text in which the selected feature does not appear, the second type positive example and the second type text A text mining program that causes a text mining device to execute an attribute value condition that is a new feature effective for classifying two types of negative examples. A text mining program in a text mining device comprising a plurality of texts, a storage device for storing attribute values for each text, and a data processing device,
A procedure for reading the text and the attribute value for each text from the storage device, and an attribute value condition which is a condition of the first type positive example and the first type negative example designated by the user, And applying the text mining to the attribute value for each text, extracting a feature effective for classifying the first type positive example and the first type negative example as a feature, and mining it in the storage device The procedure to store as a result, the procedure for causing the user to select a feature to be noted from the extracted features, and the text corresponding to the first type positive example and the first type negative example are selected. The second type positive example text in which the feature appears and the second type negative example text in which the selected feature does not appear are classified into the second type positive example and the second type negative example Generates an attribute value condition that is a new feature that is useful for Text mining program characterized by executing the steps of outputting to the output device to the data processing device. The portion that is effective for classifying the first type positive example and the first type negative example is based on the first criterion set in advance, “the appearance frequency of the first type positive example is high. The text mining program according to claim 15, wherein the phrase is a phrase having a low appearance frequency in the first type negative example text. An attribute value condition that is a new feature effective for classifying the second type positive example and the second type negative example is based on a second criterion set in advance. The text mining program according to claim 15, 16, or 17, wherein the attribute value is a combination of “attribute value having high appearance frequency as attribute value and low appearance frequency as attribute value for the second type negative example”. Text mining based on a procedure for inputting an attribute value condition which is a condition of the first type positive example and a first type negative example designated by the user, and an attribute value condition which is a condition of the first type positive example And a procedure for extracting a portion effective for classifying the first type of positive example as a feature, a procedure for causing the user to select a feature to be noticed from the features, a first type of positive example, And a procedure for separating the text corresponding to the first type negative example into the second type positive example text in which the selected feature appears and the second type negative example text in which the selected feature does not appear And a text mining device for causing a text mining device to execute a procedure for generating an attribute value condition that is a new feature effective for classifying the second type positive example and the second type negative example program. Text mining is performed based on the attribute value condition that is the condition of the first type positive example specified by the user, the text that matches the attribute value condition is set as the first type positive example, and the remaining text is set as the first type. A procedure for extracting, as a feature, a portion effective to classify as a negative example, a procedure for causing the user to select a feature to be noted from features, a first type positive example, and a first type A procedure for separating the text corresponding to the negative example into the second type of positive example text in which the selected feature appears and the second type of negative example text in which the selected feature does not appear; A text mining program that causes a text mining device to execute a procedure for generating an attribute value condition that is a new feature effective for classifying a positive example and a second type negative example. Procedures for extracting frequently occurring elements as features in all stored texts, procedures for causing the user to select features to be noted from among the features, and selection of positive text in which the selected features appear The text mining apparatus executes a procedure for generating an attribute value condition that is a new feature effective for classifying positive examples and negative examples, and separating them into negative example text that does not appear as a feature. A featured text mining program. A text mining device for extracting and outputting features from a set of texts with attributes, wherein an analysis target feature specifying means for inputting features to be noted from the features, and the input features appear in the text A positive example negative example text extraction means for extracting positive example text and negative example text from the text, and an attribute characteristic effective for classifying the positive example text and the negative example text. A text mining device comprising: an attribute feature extracting means for extracting. Text storage means for holding a set of text; attribute storage means for holding attribute values for the text; condition specifying means for inputting text mining conditions; and text mining means for extracting text features according to the conditions; Analysis target feature designation means for inputting a feature to be noted from among the features, and a positive example for extracting positive example text and negative example text from the text according to whether or not the inputted feature appears in the text A text mining apparatus comprising: negative example text extracting means; and attribute feature extracting means for extracting attribute features effective for classifying the positive example text and the negative example text. A text mining method in which a computer extracts and outputs features from a set of texts with attributes, wherein the computer inputs features to be noticed from the features, and the inputted features are included in the text. The computer extracts positive example text and negative example text from the text depending on whether it appears, and the computer has attribute features effective for classifying the positive example text and the negative example text. And a step of extracting the text mining method. The computer stores a set of texts and attribute values for the texts, the step of inputting text mining conditions to the computer, the step of extracting text features according to the conditions, and focusing on the features A step of inputting a power feature, a step of extracting positive example text and negative example text from the text according to whether or not the inputted feature appears in the text, the positive example text and the negative example text, Extracting a characteristic feature effective for classifying the text. A text mining program for causing a computer to execute a process of extracting and outputting features from a set of text with attributes, and an analysis target feature specifying process for inputting features to be noticed from among the features, and the input Effective for extracting positive example negative example text from the text and classifying the positive example text and the negative example text depending on whether the feature appears in the text A text mining program that causes the computer to execute an attribute feature extraction process for extracting a characteristic attribute. A process for storing a set of texts and attribute values for the text in a storage device, a condition designating process for inputting text mining conditions, a text mining process for extracting text features according to the conditions, and the features Analysis target feature designating process for inputting features to be focused on, and positive example negative example text extracting process for extracting positive example text and negative example text from text depending on whether or not the inputted feature appears in the text And a text mining program that causes a computer to execute attribute feature extraction processing for extracting attribute features effective for classifying the positive example text and the negative example text.

Images