JP6455068B2 - Information processing system, information processing program, and information processing method - Google Patents

Description

  The present invention relates to an information processing system, an information processing apparatus, and an information processing method that provide a user with feature words suitable for the purpose of text mining.

  Analyzing various text documents existing in the company, such as reports, questionnaires, e-mails, etc., to understand the risks and customer trends in corporate activities, and reflect them in planning business strategies and developing new products It has become important. Analyzing such characteristics and trends of text documents is called text mining. One of basic analysis techniques in text mining is a feature word extraction technique. An example of this feature word extraction technique is described in Patent Document 1. The text mining device described in Patent Document 1 appears particularly frequently in the former positive document set when the two document sets of the positive example document set and the negative example document set are given, and the latter negative example document set. A word or phrase that does not appear in the set is extracted as a feature word.

JP 2010-61176 A

  However, in the feature word extraction by this text mining device, the extracted feature words are not necessarily the features of the whole document constituting the positive example document set, but the features of a part of the document set included in the positive example document set. There is. For example, when a feature word appears in a biased manner in a document created at a specific time among documents constituting a positive document set, the feature word may represent an event peculiar to that time. . If the user performs text mining for the purpose of obtaining versatile information at any time, such feature words may not be useful. Therefore, an object of the present invention is to provide a technique for allowing a user to easily obtain a feature word suitable for the purpose of text mining.

  The first system of the present invention is a portion of feature words in a document set, wherein a difference between a feature value in the partial document set of the document set and a feature value in at least one other partial document set is a predetermined value or more. A partial feature word determining means for determining, as a partial feature word, a feature word of a document set or a feature word of a partial document set having a feature value in a partial document set of the document set equal to or greater than a predetermined value; and the partial feature word determining means And a partial feature word output means for outputting the partial feature word acquired from.

  The second system of the present invention comprises display means for displaying, out of the feature words in the document set, feature words that overlap with the partial feature words acquired from the partial feature word output means.

  The third system of the present invention comprises display means for displaying the partial feature words acquired from the partial feature word output means.

  According to the present invention, the user can easily obtain feature words suitable for the purpose of text mining.

It is a block diagram which shows the structure of 1st embodiment. It is a sequence diagram which shows operation | movement of 1st embodiment. It is an example of the document used as the object of text mining. It is explanatory drawing which shows an example of a feature word extraction result. It is explanatory drawing which shows an example of a partial feature word extraction result. It is explanatory drawing which shows an example of a partial feature word extraction result. It is an example of the display of a partial feature word. It is a block diagram which shows the structure of 2nd embodiment. It is a sequence diagram which shows operation | movement of 2nd embodiment. It is an example of the input by the user of 2nd embodiment. It is a block diagram which shows the structure of 3rd embodiment. It is a sequence diagram which shows operation | movement of 3rd embodiment. It is a sequence diagram which shows operation | movement of 3rd embodiment. It is an example of the information accumulate | stored in unnecessary attribute DB14. It is a block diagram which shows the structure of 4th embodiment. It is a block diagram which shows the hardware constitutions of 4th embodiment.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the drawings. However, the constituent elements described in the following embodiments are merely examples, and are not intended to limit the technical scope of the present invention.
(First embodiment)
<Configuration of First Embodiment>
A configuration example of the first embodiment will be described. FIG. 1 is a block diagram showing a configuration example of the first embodiment.

  The partial feature word extraction system 1 includes a partial feature word determination unit 11 and a partial feature word output unit 12. The partial feature word determination means 11 acquires from the document set DB (database) 21 a positive example document set arbitrarily designated by the user using the user terminal 3. Then, the partial feature word determination unit 11 divides the positive example document set for each attribute value attached to each document to generate a partial document set. After that, the partial feature word determination unit 11 calculates the feature degree of each partial document set for each feature word acquired from the feature word output unit 23, and determines a feature word satisfying a predetermined condition as a partial feature word. The partial feature word output unit 12 acquires the partial feature word from the partial feature word determination unit 11 and transmits it to the user terminal 3.

  The feature word extraction system 2 includes a document set DB 21, a feature word determination unit 22, and a feature word output unit 23. Each document stored in the document set DB 21 is previously assigned an attribute value. The feature word determination unit 22 acquires a positive example document set and a negative example document set arbitrarily designated by the user from the document set DB 21, and determines a feature word in the positive example document set. The feature word output means 23 acquires the feature words in the positive example document set from the feature word determination means 22 and transmits them to the partial feature word extraction system 1.

The user terminal 3 includes an input unit 31 and a display unit 32. The input unit 31 receives the designation of the positive example document set and the negative example document set from the user, and transmits the information to the partial feature word extraction system 1 and the feature word extraction system 2. The input unit 31 is an input device such as a keyboard, a mouse, or a touch panel. The user may designate the negative example document set to be transmitted to the partial feature word determination unit 11 to be different from that to be transmitted to the feature word determination unit 22. The display means 32 acquires a partial feature word from the partial feature word extraction system 1 and displays the partial feature word to the user. The display means 32 is, for example, a PC (personal computer) display, a mobile terminal display, a projection screen, or the like.
<Operation of First Embodiment>
Hereinafter, the operation of the first embodiment will be described in detail using specific examples. In this embodiment, it is assumed that the user obtains useful information for improving business results from business daily reports of business offices with good business results, and performs text mining in order to develop the information to each business office. That is, the document that is the target of feature word extraction in text mining is a daily business report collected from a plurality of sales offices. An example of the structure of the daily business report is shown in FIG. Here, the information attached to each business daily report that can be classified is an attribute value. The attribute value corresponds to information indicating, for example, “when”, “who” and “about” the document. Specifically, “August 8, 2014”, “Taro Suzuki”, “Company T”, etc. in the daily business report shown in FIG. The attribute value of each document may be registered in advance by the user in the document collection DB 21 or the like in association with each document, or when the system scans the document in advance by the user learning the positional information in the attribute value entry column. Alternatively, the attribute value may be automatically recognized from the position information and stored in association with the document. In addition, information corresponding to the superordinate concept of each attribute value is defined as an attribute category. The attribute category includes, for example, “visit date”, “creator”, “customer”, etc. in the daily business report of FIG. As with the attribute value, the attribute category may be registered in advance in the document collection DB by the user in association with each document, or may be automatically recognized and stored by the system based on the position information of the document. good. Alternatively, an attribute category corresponding to each attribute value may be stored in the DB in advance, and when the system recognizes the attribute value of the document, the attribute category may be automatically assigned from the DB or the like.

  Note that the method for classifying each document constituting the positive example document set into a partial document set is not limited to using attribute values. For example, a partial document set may be formed by collecting only even pages and collecting odd pages in the positive example document set. In that case, since the system can also suggest the relationship between rules (for example, even-numbered pages) that do not make sense to the user and the feature words, the possibility that the user can obtain the feature words suitable for the purpose increases.

  FIG. 2 is a sequence diagram illustrating an example of the operation of the first embodiment.

  First, the user designates a positive example document set and a negative example document set from the document set stored in the document set DB 21 in the feature word extraction system 2 by the input means 31 in the user terminal 3 (S1). In this embodiment, the user designates a set of daily business reports created at sales offices with good sales results as positive document sets and a set of business daily reports created at other sales offices as negative example document sets. To do. The input unit 31 transmits information specifying the two document sets designated by the user to the partial feature word extraction system 1 and the feature word extraction system 2.

  The feature word determination means 22 in the feature word extraction system 2 acquires a positive example document set and a negative example document set designated by the user from the document set DB 21 (S2). Next, the feature word determination unit 22 calculates the feature degree for words or phrases appearing in the positive example document set and the negative example document set, and determines a word or phrase satisfying a predetermined condition as a feature word (S3). The characteristic degree is an index representing how biased a certain word or phrase appears in a specific document set. For example, statistics used for testing independence such as chi-square values and log-likelihood ratios, and statistics useful for separating positive and negative document sets such as extended stochastic complexity A quantity or the like can be used as the feature degree. A predetermined condition for determining a feature word can be arbitrarily set by a user or a system designer. As the predetermined condition, for example, “the feature level is equal to or higher than a preset threshold” or “the feature level is equal to or higher than a preset rank among all words or phrases” may be set. . FIG. 4 shows an example of feature words calculated by the feature word determination means 22. For example, if the user sets the predetermined condition as “characteristic is higher than the top three of all words or phrases”, three words “new store opening”, “subsidy” and “president Takahashi” are feature words. As determined. The feature word output unit 23 acquires a feature word from the feature word determination unit 22, and transmits the acquired feature word to the partial feature word extraction system 1 (S4).

  The partial feature word determination means 11 in the partial feature word extraction system 1 acquires a positive example document set designated by the user from the document set DB 21 (S5). Then, the partial feature word determination unit 11 divides the acquired positive example document set for each attribute value attached to each document constituting the positive example document set, and generates a partial document set (S6). For example, a document set that is created by collecting only documents with the attribute value “August 8, 2014” included in the attribute category “visit date” in the positive example document set is the attribute value “2014 It is a set of partial documents with “August 8”. However, since the attribute values assigned to all the documents constituting the positive example document set cannot divide the positive example document set, the partial feature word determination unit 11 determines the attribute value from the target of the partial document set generation. It may be excluded.

  Next, the partial feature word determination unit 11 acquires a feature word for the positive example document set from the feature word output unit 23. Then, the partial feature word determining unit 11 calculates the feature degree in each partial document set for each acquired feature word (S7). As the feature degree, the same statistic as that used in the feature word determination unit 22 may be used, or a statistic used for an independence test such as a chi-square value or a log likelihood ratio, or an extended stochastic complex. Other statistics may be used in view of statistics and the like that are effective for separating a positive example document set and a negative example document set such as Kisity. However, since the calculation of the feature level is intended to identify a partial document set in which feature words appear in a biased manner, it needs to be performed in consideration of the size of each partial document set. For example, the number of appearances of feature words is divided by the size of the partial document set, and the number of appearances of feature degrees per unit document amount is used as the feature degree. This is because even if a document has a large number of appearances of feature words compared to other documents, if the number of words or phrases other than the feature words included in the entire document is also large, the feature words are included in the document set. It is because it cannot be said that it appears unevenly. The size of the set is, for example, the total number of all words or phrases included in the document set, the total number of characters included in the document set, or the like. FIG. 5 and FIG. 6 show an example of the feature degree calculated for each partial document set for each feature word. FIG. 5 shows the result of characteristic calculation for the characteristic word “new store opening”, and FIG. 6 shows the characteristic calculation for the characteristic word “President Takahashi”. For example, ID AA in the table shown in FIG. 5 indicates that the feature word “new store opening” is calculated as “5.3” in the partial document set with the attribute value “Taro Konishi”.

  Subsequently, the partial feature word determining unit 11 refers to the calculated feature degree and determines a feature word satisfying a predetermined condition as a partial feature word (S8, S9). The predetermined condition is a condition for separating a feature word that appears biased in a partial document set and a feature word that does not. The predetermined condition can be arbitrarily set by the user or the system designer, and may be the same as the condition used when the feature word determining means 22 determines the feature word, or may be another condition. Other conditions include, for example, “more than the smallest value of the feature words determined by the feature word determination unit 22” or “other feature levels calculated by the partial feature word determination unit 11”. 3 or more "or" 5 or more times larger than any of the other feature degrees calculated by the partial feature word determination means 11 ". Here, it is assumed that the user has set the condition “more than the smallest value among the features of the feature words determined by the feature word determination unit 22”. As shown in FIG. 4, among the feature words determined by the feature word determination means 22, the smallest is “7.8” of “President Takahashi”. Therefore, the partial feature word determining means 11 determines a feature word that is calculated with “feature degree of 7.8 or more” in the specific partial document set as a partial feature word. In the table shown in FIG. 5, since there is no partial document set for which the feature degree satisfying the condition is calculated, the partial feature word determination unit 11 does not determine the feature word “new store opening” as a partial feature word. On the other hand, with the ID BA in the table shown in FIG. 6, the characteristic degree “42.0” that satisfies the condition is calculated. That is, from FIG. 6, the feature word “President Takahashi” is calculated as “42.0” in the partial document set with the attribute value “Company T”. Accordingly, since the condition “feature degree is 7.8 or higher” is satisfied, the partial feature word determination unit 11 determines the feature word “President Takahashi” as the partial feature word.

  A feature word in the positive example document set is also characteristic in a part of the document set constituting the positive example document set. This means that the feature word appears biased in the partial document set. means. Therefore, the partial feature word determination unit 11 determines a feature word that appears biased in a specific partial document set as a partial feature word.

  The partial feature word output unit 12 acquires a partial feature word from the partial feature word determination unit 11, and transmits the acquired partial feature word to the user terminal 3 (S10).

  The display means 32 in the user terminal 3 displays the partial feature word for the user (S11). The display means 32 may display the attribute value attached to the partial document set in which the partial feature word appears characteristically together with the partial feature word. For example, when “President Takahashi” is determined as a partial feature word as in the above-described specific example, the display means 32 may display the attribute value “Company T” together with “President Takahashi” to the user. Thus, the user can know that the word “President Takahashi” that appears characteristically in the positive example document set appears biased in the document set with the attribute value “T company”. . Furthermore, the display means 32 may display the attribute category about the attribute value attached | subjected to the partial document set in which a partial feature word characteristically appears with a partial feature word. In the specific example described above, “customer” corresponds to the attribute category for the attribute value “Company T”. An example of the display of the partial feature words by the display means 32 is shown in FIG. The user who sees the display in FIG. 7 calculates the feature word “42.0” in the partial document set to which the feature word “President Takahashi” has the attribute value “Company T” included in the attribute category “business partner”. It is possible to know that it appears biased in the partial document set.

  Finally, the display means 32 displays to the user the feature words that do not overlap with the partial feature words among the feature words in the positive example document set (S12). If there is no feature word satisfying the predetermined condition in S8, the display means 32 displays all the feature words in the positive example document set.

  Moreover, the display means 32 may display all feature words including a feature word overlapping with the partial feature word at the same time to the user. In that case, the feature word display means 32 can distinguish between feature words that overlap with partial feature words and feature words that do not, such as displaying them as separate tables or displaying them with different marks on the same table. To the user.

  As a result, the user should not appear to be a feature word that “Takahashi President” appears in the collection of positive examples, but rather appear in a document about “Company T”, which is one of the “business partners”. Can know. And because of the relationship between the characteristic word “President Takahashi”, the attribute category “Business Partner”, and the attribute value “Company T”, the user has “President Takahashi” as individual information for a specific business partner and is not suitable for the purpose. It can be determined that it is a feature word.

The display means 32 may display only feature words that do not overlap with the partial feature words to the user. That is, the partial feature words described in S11 may not be displayed, and only the feature words that do not overlap with the partial feature words may be displayed as described in S12. As a result, when the user regards feature words that appear without deviation in the collection of positive example documents as useful information, the user can obtain only feature words that meet the purpose of text mining from the beginning.
(Second embodiment)
<Configuration of Second Embodiment>
Next, a configuration example of the second embodiment will be described. FIG. 8 is a block diagram showing a configuration example of the second embodiment. The system according to the present embodiment is different from the first embodiment in that the feature word displayed on the user by the display unit 32 changes depending on the input from the input unit 31. The input unit 31 receives an input of a feature word that the user has determined to be hidden, and transmits the feature word to the display unit 32. The display unit 32 displays a feature word that does not overlap with the feature word acquired from the input unit 31 among the feature words acquired from the feature word output unit 23 to the user. Since other configurations and operations are the same as those in the first embodiment, descriptions other than the input unit 31 and the display unit 32 are omitted.
<Operation of Second Embodiment>
FIG. 9 is a sequence diagram illustrating an example of the operation of the second embodiment. In FIG. 9, S1 to S11 are the same as those in FIG.

  The user refers to the display of the partial feature words by the display means 32, and inputs the feature words determined to be hidden to the input means 31 (S22). To explain using a specific example similar to the first embodiment, the user receives the display of the attribute value “Company T” and the attribute category “business partner” from the display unit 32 together with the partial feature word “President Takahashi”. The user who received the display judges the relationship between “President Takahashi”, “Company T”, and “Customer”, and “President Takahashi” is information that represents individual circumstances in the attribute value “T Company”. Judge that it is not a useful feature word. This is because information that is specific to a specific business partner is likely to be information that does not contribute to the improvement of sales results at a sales office that is not related to the business partner, and is suitable for the purpose of the user in this specific example. This is because there is a high possibility that it is not. Then, the user inputs the feature word “President Takahashi” to the input means 31. The input unit 31 transmits the input feature word to the display unit 32.

  An example of display by the display means 32 and acceptance of input by the input means 31 is shown in FIG. The user who has received the partial feature word display as shown in FIG. 10 from the display means 32 uses the input means 31 and checks the check box provided at the right end of the line indicating “President Takahashi” to be hidden. Enter. Thereby, the input means 31 receives the input of the feature word “President Takahashi” from the user.

  The display unit 32 extracts a feature word that overlaps with the feature word acquired from the input unit 31 from the feature words acquired from the feature word output unit 23 (S23). Then, the display unit 32 displays the feature words without duplication to the user (S24).

In the second embodiment, the user can determine the usefulness of a partial feature word from the attribute value or attribute category in which the partial feature word appears characteristically, and the determination result can be reflected in the result of feature word extraction. That is, the feature words that are finally displayed to the user reflect the user's intention more than in the first embodiment.
(Third embodiment)
<Configuration of Third Embodiment>
Next, a configuration example of the third embodiment will be described. FIG. 11 is a block diagram illustrating a configuration example of the third embodiment. The system according to the present embodiment is different from the second embodiment in that it includes a non-display attribute determining unit 13, a non-display attribute DB 14, and a display control unit 33.

  The non-display attribute determining unit 13 acquires the calculation result of the feature degree in each subset of each feature word from the partial feature word output unit 11, and acquires the feature word input by the user from the input unit 31. Then, the non-display attribute determining unit 13 determines an attribute value assigned to the partial document set in which the feature word input by the user appears characteristically as the non-display attribute value. The partial document set in which a feature word appears characteristically is a partial document set calculated with a value satisfying a predetermined condition for the feature degree of the feature word. The predetermined condition is a condition used for determining the partial feature word in the partial feature word determining means 11 described above. An additional condition may be added to the condition for determining the non-display attribute in the non-display attribute determining unit 13. For example, when a user inputs a plurality of feature words, the attribute values that are commonly assigned to the partial document sets in which those feature words appear characteristically or the attribute values that are assigned most frequently are The display attribute determining means 13 may determine the non-display attribute value. The non-display attribute determining unit 13 transmits the non-display attribute to the non-display attribute DB 14 and accumulates it.

  The non-display attribute determining unit 13 may weight the non-display attribute value. The weighting to each non-display attribute value may be a numerical value arbitrarily set by the user, for example, or may be a numerical value corresponding to the number of times each attribute value is determined as the non-display attribute value. In the latter case, each time an attribute value is determined as a non-display attribute value by the non-display attribute determining means 13, the weighting of the attribute value stored in the non-display attribute DB 14 may be updated. The weighting to each non-display attribute value is referred to in processing in the partial feature word display means 11 and the feature word display means 33 described later.

  The display control unit 33 refers to the non-display attribute DB 14 and controls the display unit 32 not to display characteristic feature words in the partial document set to which the attribute values stored in the non-display attribute DB 14 are attached. As a result, characteristic feature words in the partial document set with the non-display attribute value are not displayed to the user. When the non-display attribute value is weighted, the display control unit 33 refers to the weighted attribute value only when the weight satisfies a predetermined condition. The predetermined condition can be arbitrarily set by the user or the system designer. For example, “the weight is 1 or more” can be set. In this case, the display control unit 33 refers to only the non-display attribute value having “weighting of 1 or more”, and controls the display unit 32 not to display characteristic feature words in the partial document set with the attribute value. To do.

Since other configurations and operations are the same as those in the second embodiment, descriptions other than the non-display attribute determining unit 13, the non-display attribute DB 14, and the display control unit 33 are omitted.
<Operation of Third Embodiment>
12 and 13 are sequence diagrams illustrating an example of the operation of the third embodiment.

  FIG. 12 is a sequence diagram showing processing until the non-display feature word input by the user is registered in the non-display attribute DB 14. S11 and S22 shown in FIG. 12 are the same as those in FIG.

  The partial feature word output means 11 transmits the calculation result of the feature degree in each subset of each feature word to the non-display attribute determination means 13 (S31). The non-display attribute determining unit 13 obtains the calculation result of the degree of feature in each subset of each feature word and the feature word input by the user (S32). Subsequently, the non-display attribute determining unit 13 determines an attribute value assigned to the partial document set in which the feature word input by the user appears characteristically as a non-display attribute value (S33). Specifically, the non-display attribute determining unit 13 refers to the calculation result of the feature value in each subset of each feature word, and the feature value for the feature word input by the user is calculated with a value satisfying a predetermined condition. If there is a partial document set, the attribute value assigned to the partial set is determined as a non-display attribute value. To explain with a specific example similar to that of the above-described embodiment, the non-display attribute determination unit 13 includes the attribute value “T” attached to the partial document set in which the non-display feature word “President Takahashi” input by the user appears characteristically. "Company" is determined as the non-display attribute value. Then, the non-display attribute determining means 13 transmits the non-display attribute value to the non-display attribute DB 14 and accumulates it (S34). Note that the non-display attribute determining unit 13 may weight the non-display attribute value. In this case, the non-display attribute is stored in the non-display attribute DB 14 together with the non-display attribute value and the weight associated with each non-display attribute value.

  When the user does not input a feature word, no new non-display attribute value is accumulated in the non-display attribute DB 14.

  Further, the non-display attribute determining unit 13 may accumulate the attribute category in the non-display attribute DB. To explain with the above-described specific example, the non-display attribute determination means 13 sets “customer” as the attribute category of the attribute value attached to the partial document set in which the non-display feature word “President Takahashi” appears characteristically, It may be determined as a non-display attribute category and stored in the non-display attribute DB 14. An example of information stored in the non-display attribute DB 14 is shown in FIG. In FIG. 14, the non-display attribute category “business partner” is stored in the non-display attribute DB 14 together with the weight “1.0”.

  FIG. 13 shows a sequence diagram when a new feature word extraction process is started again after the process shown in FIG. 12 is completed. In FIG. 13, S1 to S22 are the same as those in FIG.

  The display control unit 33 refers to the non-display attribute DB 14 and controls the display unit 32 not to display characteristic partial feature words in the partial document set to which the non-display attribute value is assigned. Thereby, the display means 32 displays a partial feature word other than a characteristic partial feature word in the partial document set with the non-display attribute value to the user (S40). When the non-display attribute category is accumulated in the non-display attribute DB 14, the display control unit 33 displays the characteristic partial feature words in the partial document set to which the attribute value included in the non-display attribute category is added. Control not to display. When the non-display attribute value or the non-display attribute category is weighted, the display control unit 33 controls the display unit 32 with reference to only the attribute value and the attribute category whose weight satisfies a predetermined condition. Specifically, for example, the display control unit 33 acquires data as shown in FIGS. 5 and 6 from the partial feature word output unit 12 and acquires data as shown in FIG. 14 from the non-display attribute DB 14. Then, the display control means 33 refers to the non-display attribute category shown in FIG. 14 whose weight satisfies a predetermined condition. If the user sets the predetermined condition as “attribute category with weighting of 0.5 or more”, the display control means 33 refers to the attribute categories “business partner” and “person in charge” from FIG. Subsequently, the display control means 33 refers to the feature value calculation data for each feature word shown in FIG. 5 and FIG. 6 and is assigned the attribute values included in the attribute categories “business partner” and “person in charge”. In the document set, it is determined whether there is a feature word whose feature degree satisfies a predetermined condition. If the user sets the predetermined condition as “feature degree is 5.0 or higher”, the characteristic word “new store” is assigned the attribute value “Taro Konishi” included in the attribute category “person in charge” from FIG. In the partial document set, the feature degree satisfies the condition. As shown in FIG. 6, the feature word “President Takahashi” satisfies the feature condition in the partial document set to which the attribute value “Company T” included in the attribute category “customer” is attached. Therefore, the display control means 33 controls the display means 32 not to display both the characteristic words “new store opening” and “president Takahashi”.

  In addition, when the user inputs a feature word, the display control unit 33 performs control so that the input feature word is not displayed on the display unit 32 (S41).

  Further, the display control means 33 refers to the non-display attribute DB 14 and controls the display means 32 not to display characteristic feature words in the partial document set to which the attribute values stored in the non-display attribute DB 14 are attached. (S42). When the non-display attribute category is accumulated in the non-display attribute DB 14, the display control unit 33 displays the characteristic partial feature words in the partial document set to which the attribute value included in the non-display attribute category is added. Control not to display. When the non-display attribute value or the non-display attribute category is weighted, the display control unit 33 controls the display unit 32 with reference to only the attribute value and the attribute category whose weight satisfies a predetermined condition. Thereby, the feature words finally displayed to the user are the feature words that have not been excluded from the display target in S41 and S42.

In the third embodiment, information about non-display feature words input by the user can be fed back to a new feature word extraction process after the next time. In other words, the system stores attribute values and attribute categories in which feature words that the user has determined to be unsuitable for use appear unevenly, and the feature words that are characteristic for those attribute values are automatically removed in the subsequent processing. Can do. For example, if a characteristic word that is characteristic for a specific business partner “Company T” is unnecessary, the user sets the attribute value “Company T” as a hidden attribute value, and a characteristic characteristic for any business partner. If the target category is unnecessary, the attribute category “business partner” may be set to the non-display attribute category. Thereby, in this embodiment, the feature word reflecting a user's intention can be provided by simpler operation than 2nd embodiment.
(Fourth embodiment)
<Configuration of Fourth Embodiment>
A configuration example of the fourth embodiment will be described. FIG. 15 is a block diagram showing a configuration example of the fourth embodiment.

  The partial feature word extraction system 1 includes a partial feature word determination unit 11 and a partial feature word output unit 12. The partial feature word determining unit 11 divides the document set for each attribute value attached to the document constituting the document set, and generates a partial document set. Thereafter, the partial feature word determination unit 11 acquires the feature words in the document set, and calculates the feature degree of the feature word for each partial document set. The partial feature word determination unit 11 determines a feature word as a partial feature word when there is a feature word for which a characteristic degree satisfying a predetermined condition is calculated in a specific partial document set. The predetermined condition is, for example, “a difference is 3 or larger compared to any of the other feature degrees calculated by the partial feature word determining unit 11” and “of other feature degrees calculated by the partial feature word determining unit 11 The difference is 5 or more compared to at least one of them. ” The partial feature word output unit 12 acquires the partial feature word from the partial feature word determination unit 11 and outputs it. The partial feature word determining unit 11 and the partial feature word output unit 12 may be realized by separate hardware devices.

  This allows the user to distinguish feature words that appear biased in some of the documents that constitute the positive document set from feature words that do not, and thus easily obtain feature words suitable for the purpose of text mining. be able to.

  Next, the hardware configuration of the fourth embodiment will be described with reference to FIG. The CPU 41 is a processor for arithmetic control, and implements each functional unit of the present invention by executing a program. The ROM 42 stores fixed data and programs such as initial data and programs. The communication control unit 43 communicates with an external device via a network. The partial feature word extraction system 1 exchanges information with a PC, a user terminal, and other systems used by the user via the communication control unit 43. The RAM 44 is a random access memory that the CPU 41 uses as a work area for temporary storage. The RAM 44 has a capacity for storing various data necessary for realizing the present embodiment. The storage unit 45 stores data such as a database necessary for realizing the present embodiment and an application program executed by the CPU 41 in a nonvolatile manner.

Moreover, transmission / reception of information between each system and the terminal demonstrated in each embodiment is performed via a communication network. Communication may be wireless or wired. The communication network between each system and terminal may be the same network such as the Internet. Each of the above-described embodiments is merely an example embodying the present invention, and within the scope of the gist of the present invention described in the claims. If there are, various changes can be made.

DESCRIPTION OF SYMBOLS 1 Partial feature word extraction system 2 Feature word extraction system 3 User terminal 11 Partial feature word determination means 12 Partial feature word output means 13 Non-display attribute determination means 14 Non-display attribute DB
21 Document collection DB
22 feature word determination means 23 feature word output means 31 input means 32 display means 33 display control means 41 CPU
42 ROM
43 Communication control unit 44 RAM
45 storage unit

Claims (10)

Among the feature words in the document set,
A feature word of a partial document set in which a difference between a feature value in the partial document set of the document set and a feature value in at least one other partial document set is a predetermined value or more, or a feature value in the partial document set of the document set A partial feature word determining means for determining, as a partial feature word, a feature word of a partial document set whose is equal to or greater than a predetermined value;
Partial feature word output means for outputting the partial feature word acquired from the partial feature word determination means;
An information processing system comprising: The information processing system according to claim 1, further comprising: a display unit configured to display a feature word in the document set other than the feature word overlapping the partial feature word acquired from the partial feature word output unit. The information processing system according to claim 1, further comprising display means for displaying the partial feature words acquired from the partial feature word output means. The information processing system according to any one of claims 1 to 3, wherein the partial document set is generated by collecting documents having common attribute values among documents constituting the document set. The information processing unit according to claim 4, wherein the display unit displays an attribute value commonly assigned to documents forming a partial document set in which the partial feature word appears characteristically or an attribute category for classifying the attribute value. system. An input means for a user to input an arbitrary feature word is provided,
The information processing system according to claim 2, wherein the display unit displays a feature word other than the feature word that overlaps the feature word acquired from the input unit among the feature words in the document set. Non-display for determining, as a non-display attribute value, an attribute value satisfying a predetermined condition among attribute values commonly assigned to documents that form a partial document set in which a feature word acquired from the input unit appears characteristically Attribute determination means;
A hidden attribute database for registering the hidden attribute value;
The display unit displays, among feature words in the document set, features other than feature words that appear characteristically in a partial document set composed of documents with attribute values registered in the non-display attribute database. 6. The information processing system according to 6. Among the attribute categories in which attribute values commonly assigned to the documents that form a partial document set in which the feature words acquired from the input unit characteristically appear are classified, the attribute categories that satisfy a predetermined condition are not displayed. A non-display attribute determining means for determining as a category,
A hidden attribute database for registering the hidden attribute category;
The display means characteristically appears in a partial document set composed of documents with attribute values classified into a non-display attribute category registered in the non-display attribute database among characteristic words in the document set. The information processing system according to claim 6 or 7, wherein information other than the characteristic words is displayed. Among the feature words in the document set,
A feature word of a partial document set in which a difference between a feature value in the partial document set of the document set and a feature value in at least one other partial document set is a predetermined value or more, or a feature value in the partial document set of the document set A partial feature word determination process for determining, as a partial feature word, a feature word of a partial document set whose is equal to or greater than a predetermined value;
A partial feature word output process for outputting the partial feature word acquired from the partial feature word determination means;
An information processing program that causes a computer to execute. Computer
Among the feature words in the document set,
A feature word of a partial document set in which a difference between a feature value in the partial document set of the document set and a feature value in at least one other partial document set is a predetermined value or more, or a feature value in the partial document set of the document set Determine a feature word of a partial document set whose is greater than or equal to a predetermined value as a partial feature word,
Outputting the determined partial feature word;
Information processing method.

Images