JP2014232494A - Document creation assist device and operation method thereof - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain a prediction value of a response to a document.SOLUTION: A response prediction acquisition unit 4 searches a response history DB for a record fulfilling (condition 1) that a difference between a time of day 202 and the current time of day is within a prescribed range, (condition 2) that a situation of whether or not the day in question is a holiday matches a holiday flag 203, (condition 3) that the number of characters in a document concerned matches the number of characters 206, and (condition 4) that a category of the document concerned and category information 204 match (S2071). The response prediction acquisition unit 4 reads out a first period response count 211, a second period response count 212, a third period response count 213, a fourth period response count 214, and a fifth period response count 215, respectively as a first period response prediction value, a second period response prediction value, a third period response prediction value, a fourth period response prediction value, and a fifth period response prediction value (S2077), and then terminates the process. A response prediction output unit 5 outputs the first period response prediction value, the second period response prediction value, the third period response prediction value, the fourth period response prediction value, and the fifth period response prediction value.

Description

  The present invention relates to a document creation support apparatus and an operation method thereof.

  Conventionally, a document is created by the creator himself / herself, or with the assistance of saving the creator's effort based on a predetermined rule.

Japanese Patent Laid-Open No. 02-297124 JP 02-116956 A

  When a creator creates a document by himself / herself, he / she needs to be proficient in creating a document that easily obtains a reaction by posting the document and having a person see it and getting a response.

  On the other hand, when a document is created with assistance, the creation time is shortened, but it is not known whether the document is easy to get a response until the response is actually obtained.

  In any case, if the creator is not accustomed to document creation, a document that easily obtains a response cannot be created or a response cannot be predicted.

  The present invention has been made in view of the above problems, and an object of the present invention is to provide a document creation support apparatus capable of obtaining a predicted value of reaction to a document and an operation method thereof. .

  In order to solve the above-described problem, a document creation support apparatus according to a first aspect of the present invention obtains a feature quantity of a target document that is a support target of document creation, and a reaction prediction that is a predicted value of a response to the target document. A reaction prediction acquisition unit for obtaining a value based on the feature amount is provided.

  For example, the document creation support apparatus includes a plurality of records for a document that is not the target document, and each record includes a database that includes a feature amount of the document and the number of responses to the document, and the response prediction acquisition The unit searches the database for a record corresponding to the feature quantity of the target document, and obtains the predicted response value based on the number of reactions included in the record.

  For example, the feature amount includes the number of characters after unnecessary characters are deleted from the target document.

  For example, the feature amount includes the number of characters after unnecessary characters are deleted from the target document, the category of the target document, and the subcategory of the target document.

  For example, the document creation support apparatus includes a record for each group including a plurality of documents having a common category and subcategory, and the record includes a document classification database including elements indicating characteristics of the group, The reaction prediction acquisition unit generates a feature vector indicating the feature of the target document, searches a record corresponding to the target document category from the document classification database, and for each searched record, from an element included in the record The representative point vector is generated, the distance between the feature vector and each representative point vector is obtained, and the subcategory corresponding to the minimum distance is set as the subcategory of the target document.

  A document creation support apparatus according to a second aspect of the present invention provides a document input unit for inputting a target document, a feature amount of the target document, and a predicted response value that is a predicted value of a response to the target document. And a reaction prediction output unit for outputting the predicted reaction value.

  According to a third aspect of the present invention, there is provided a method for operating a document creation support apparatus, in which a response prediction acquisition unit of a document creation support apparatus obtains a feature amount of a target document, and the reaction prediction acquisition unit responds to the target document. And a step of obtaining a predicted reaction value, which is a predicted value of the above, based on the feature amount.

  According to a fourth aspect of the present invention, there is provided a document creation support apparatus operating method in which a document input unit of a document creation support apparatus inputs a target document that is a document creation support target, and a response prediction acquisition of the document creation support apparatus. A step of obtaining a feature amount of the target document; a step of obtaining a response prediction value that is a predicted value of a response to the target document based on the feature amount; and the document creation support. The reaction prediction output unit of the apparatus includes a step of outputting the reaction prediction value.

  According to the present invention, a predicted response value that is a predicted value of a response to a document can be obtained.

It is a figure which shows the structure of the document preparation assistance apparatus which concerns on 1st Embodiment. It is a figure which shows an example of a data structure of reaction log | history DB2. It is a figure which shows the example of a screen of the display apparatus connected to the document preparation assistance apparatus. 4 is a flowchart showing a process flow of a document input unit 1. 4 is an overall flowchart showing a flow of processing of a reaction prediction acquisition unit 4 and a reaction prediction output unit 5. It is a flowchart of acquisition of a reaction predicted value by the reaction prediction acquisition unit 4 (S207). It is a figure which shows an example of the record searched by the reaction prediction acquisition part 4 from reaction history DB2. It is a figure which shows another example of a screen of the display apparatus connected to the document preparation assistance apparatus. It is a figure which shows the structure of the document preparation assistance apparatus which concerns on 2nd Embodiment. It is a figure which shows an example of a data structure of reaction prediction DB6. It is a figure which shows an example of a data structure of document classification DB7. It is a figure which shows an example of a data structure of topic DB8. It is a flowchart which shows the flow of the production | generation process of document classification DB7 by the reaction prediction acquisition part. It is a figure which shows the example of the past document vector. 4 is an overall flowchart showing a flow of processing of a reaction prediction acquisition unit 4 and a reaction prediction output unit 5. It is a flowchart of target document subcategory detection (S205) by the reaction prediction acquisition unit 4. It is a flowchart of acquisition (S208) of the reaction predicted value by the reaction prediction acquisition part 4. FIG.

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

[First Embodiment]
FIG. 1 is a diagram illustrating a configuration of a document creation support apparatus according to the first embodiment.

  The document creation support apparatus includes a document input unit 1 that inputs a document (target document) that is a document creation support target, and a reaction history database (a history of reactions to documents created in the past (past documents)). (Hereinafter, the database is abbreviated as DB) DB2, the reaction acquisition unit 3 that acquires the response to the document, the feature quantity of the target document is obtained, and the response predicted value that is the predicted value of the response to the target document is used as the feature quantity. The reaction prediction acquisition part 4 calculated | required based on and the reaction prediction output part 5 which outputs a reaction prediction value are provided.

  FIG. 2 is a diagram showing an example of the data structure of the reaction history DB 2 and shows a portion of the category “Japanese movie”.

  The reaction history DB 2 has a record for each document generated in the past (hereinafter referred to as a past document).

  Each record includes a document number 201 of the past document, a time 202 when the past document was generated, a holiday flag 203 indicating whether or not the date when the past document was generated, a category information 204 of the past document, and a title 205 of the past document. , The number of characters 206 of the past document, the past document 207 that is the past document itself, the first period of the first period, the second period, the third period, the fourth period, and the fifth period that sequentially arrive after the past document is created The number of reactions in the first period 211, the number of reactions in the second period 212, the number of reactions in the second period 212, the number of reactions in the third period 213, the number of reactions in the third period 213, the number of reactions in the fourth period A fourth period reaction number 214, which is the number of reactions, and a fifth period reaction number 215, which is the number of reactions in the fifth period, are provided.

  The first period is, for example, 10 minutes after the creation, the second period is 10 minutes to 1 hour later, the third period is 1 hour to 2 hours later, the fourth period is 2 hours to 5 hours later Until later, the fifth period is defined as from 5 hours to 10 hours later. A period may be further provided.

  Note that the items to be stored in the record of the reaction history DB 2 are not limited to this, and the degree of specialization of the document is calculated and digitized based on the kanji and hiragana ratio in the document and the dictionary of word specialities created in advance. Alternatively, the document politeness based on a pre-created ending dictionary or the document attention calculated on the basis of a pre-created word attention dictionary may be used.

  For each actual document (past document), the response acquisition unit 3 counts the number of responses in each of the first, second, third, fourth, and fifth periods that arrive sequentially after the document is created. Are stored in the first period reaction number 211, the second period reaction number 212, the third period reaction number 213, the fourth period reaction number 214, and the fifth period reaction number 215 of the corresponding record in the reaction history DB2. For example, the reaction acquisition unit 3 updates the reaction history DB 2 every time a new reaction is obtained.

  As a response to a document, the following indices expressed numerically can be used. That is, the number of responses is (1) if the document is a question message on a question board, for example, the number of obtained answers, (2) if the document is an email, for example, the number of replies, and (3) the document is a social networking service. If it is a posted article of (social networking service, SNS), for example, it is the number of evaluation button presses and the number of comments. (4) If the document is a posted article on a web page, for example, it is the number of browsing, the number of evaluation button stamps, and the number of comments.

  For example, when a document is displayed on a computer screen used by a friend of the author and the friend presses the evaluation button on the screen, the reaction acquisition unit 3 acquires the number of times of pressing as the number of reactions, and updates the reaction acquisition unit 3 .

  FIG. 3 is a diagram illustrating a screen example of the display device connected to the document creation support device.

  The document input unit 1 displays the document input screen 100 on the display device, and the reaction prediction output unit 5 displays the reaction prediction screen 200. When a document is input to the document input screen 100, the reaction prediction output unit 5 displays a reaction prediction value on the reaction prediction screen 200. When the posting button 300 is pressed, the document is transmitted (posted) to the outside.

  For a document that has been created in this way, that is, a past document, a record relating to the document is generated in the reaction history DB 2.

  The display contents of the reaction prediction screen 200 will be described later.

  FIG. 4 is a flowchart showing a process flow of the document input unit 1.

  The document input unit 1 determines whether a document has been input to the document input screen 100 or whether the content of the document has changed (S101). If the document has been input or the content of the document has changed (S101) (S101: YES), the reaction prediction acquisition unit 4 and the reaction prediction output unit 5 are called (S103), and the process returns to step S101. The document creator selects, for example, a document category from a plurality of categories, the category is given to the document input unit 1, and the document input unit 1 calls the reaction prediction acquisition unit 4. A document in the document input screen 100 (hereinafter referred to as a target document) and the category (hereinafter referred to as a target document category) are given to the reaction prediction acquisition unit 4.

  FIG. 5 is an overall flowchart showing a processing flow of the reaction prediction acquisition unit 4 and the reaction prediction output unit 5.

  The reaction prediction acquisition unit 4 decomposes the target document into words (S201), deletes unnecessary words (S203), detects the number of characters of all remaining words (hereinafter referred to as word groups) (S204), and reacts. A predicted value is acquired (S207).

  Next, the reaction prediction output unit 5 displays the reaction prediction value on the reaction prediction screen 200 (S209), and the control returns to S101.

  The predicted response value is not limited to display, and may be printed.

  For example, the reaction prediction acquisition unit 4 holds in advance an unnecessary word list that covers unnecessary words, and in S203, searches the target document for words and punctuation marks that match the words in the unnecessary word list and deletes them.

  For example, in S204, the reaction prediction acquisition unit 4 rounds off the last digit of the number of characters, and sets this as the number of characters of the target document. This is to prevent the number of corresponding records from decreasing in the search of the reaction history DB 2 described later.

  FIG. 6 is a flowchart of obtaining a predicted reaction value (S207) by the reaction prediction obtaining unit 4.

  The reaction prediction acquisition unit 4 matches the holiday flag 203 with whether or not the difference between the (time 1) time 202 and the current time is within a predetermined range from the reaction history DB 2 (condition 2) whether the current day is a holiday. A record satisfying (condition 3) that the number of characters of the target document matches the number of characters 206 and (condition 4) that the target document category matches the category information 204 is searched (S2071).

  FIG. 7 is a diagram illustrating an example of a record retrieved from the reaction history DB 2 by the reaction prediction acquisition unit 4.

  For example, when the current time is “10:01”, the current day is not a holiday, the number of characters of the target document is “200”, and the category of the target document is “Japanese movie”, for example, time 202 “10:00”, holiday flag 203 Three records including “0”, the number of characters 206 “200”, and the category information 204 “Japanese movie” are searched.

  Returning to FIG. 6, if the number of records is 2 or more (S2073: YES), the reaction prediction acquisition unit 4 averages the first period reaction number 211, the second period reaction number 212, and the third period reaction number 213. The average of the fourth period reaction number 214 and the average of the fifth period reaction number 215 is calculated, and the average of the first period reaction number 211 is defined as a predicted reaction value in the first period (hereinafter referred to as the first period response prediction value). The average of the second period response number 212 is defined as a predicted response value in the second period (hereinafter referred to as second period response predicted value), and the average of the third period response number 213 is defined as a predicted response value in the third period (hereinafter referred to as the third period). Reaction predicted value), the average of the fourth period response number 214 is the response predicted value in the fourth period (hereinafter referred to as fourth period response predicted value), and the average of the fifth period response number 215 is the predicted response value in the fifth period. (Hereinafter, the fifth period response prediction value) ( 2075), reads out respectively, the process ends.

  Taking FIG. 7 as an example, the first period response predicted value is “2”, the second period response predicted value is “1”, the third period response predicted value is “1”, and the fourth period response predicted value is “1”. ”, The fifth period response prediction value is“ 0.66 ”.

  If the number of records is 1 (S2073: NO), the reaction prediction acquisition unit 4 sets the first period reaction number 211 as the first period reaction prediction value, the second period reaction number 212 as the second period reaction prediction value, and the third period reaction. The number 213 is the third period response predicted value, the fourth period response number 214 is the fourth period response predicted value, and the fifth period response number 215 is the fifth period response predicted value (S2077). .

  The reaction prediction output unit 5 takes the time P on the horizontal axis and the reaction prediction value on the vertical axis on the reaction prediction screen 200 in FIG. 3, and points P1 and P1 corresponding to the first period reaction prediction value and the first period end time. The point P2, corresponding to the sum of the 1-period response prediction value and the second-period response prediction value and the second period end time, the sum of the first-period response prediction value, the second-period response prediction value, and the third-period response prediction value Corresponds to point P3 corresponding to the third period end time, first period response predicted value, second period response predicted value, third period response predicted value, and sum of fourth period response predicted value and fourth period end time Corresponding to the sum of the point P4, the first period response predicted value, the second period response predicted value, the third period response predicted value, the fourth period response predicted value and the fifth period response predicted value, and the fifth period end time. A line graph connecting the points P5 is displayed.

  Thereby, the author of the target document can know the transition of the cumulative reaction number when the current document is posted. In addition, you may display a reaction predicted value with a number etc. The predicted reaction value may be printed.

(Modification)
Or in this Embodiment, you may employ | adopt the following aspects.

  In step S2071, the reaction prediction acquisition unit 4 temporarily sets the number of characters of the target document to 50, 100, 150, 200, 250.

  As shown in FIG. 8, the reaction prediction output unit 5 takes the number of characters on the horizontal axis, the reaction prediction value (for example, the first period reaction prediction value) on the vertical axis, and the reaction prediction value and the number of characters when the number of characters is 50. Corresponds to the point P11 corresponding to “50”, the response predicted value when the number of characters is 100 and the point P12 corresponding to the number of characters “100”, the predicted response value and the number of characters “150” when the number of characters is 150 A point P13 corresponding to the response predicted value when the number of characters is 200 and the number of characters “200” and a point P15 corresponding to the response predicted value and the number of characters “250” when the number of characters is 250 are connected. A broken line graph is displayed, and a point P12 corresponding to the number of characters of the target document, for example, the number of characters “100” is highlighted.

  As a result, the author of the target document can obtain knowledge that how much the number of characters in the current document can be increased or decreased to increase the number of responses.

  In addition, the number of characters (optimum number of characters) having the highest response predicted value may be obtained, and the current number of characters and the optimum number of characters may be displayed as numbers. The predicted reaction value may be printed.

[Second Embodiment]
Next, a second embodiment of the present invention will be described. In the second embodiment, the same or similar apparatus and apparatus configuration as those in the first embodiment are used, and the same or similar elements are redundantly described by using the reference numerals used in the first embodiment. For brevity, the description will focus on matters different from the first embodiment.

  FIG. 9 is a diagram illustrating a configuration of a document creation support apparatus according to the second embodiment.

  The document creation support apparatus includes a document input unit 1, a reaction history DB 2, a reaction acquisition unit 3, a reaction prediction acquisition unit 4, a reaction prediction output unit 5, and a reaction prediction DB 6 in which data for reaction prediction is stored. And a document classification DB 7 for storing data obtained by classifying documents, and a topic DB 8 for storing data related to the topic of the document.

  FIG. 10 is a diagram illustrating an example of a data configuration of the reaction prediction DB 6.

  The reaction prediction DB 6 has a common value for each of the time 202, holiday flag 203, category information 204, and number of characters 206 in the reaction history DB 2, and one or more common sub-categories of the category information 204 are further subdivided. When there are a plurality of documents (referred to as groups), each group has a record.

  Each record has the same time 601 as the corresponding time 202, the same holiday flag 602 as the corresponding holiday flag 203, the same category information 603 as the corresponding category information 204, the subcategory information 604 indicating the corresponding subcategory, and the same number of characters 206. From the number of characters 605, the first period reaction number 611 calculated from the corresponding first period reaction number 211, the second period reaction number 612 calculated from the corresponding second period reaction number 212, and the corresponding third period reaction number 213 A third period reaction number 613 calculated, a fourth period reaction number 614 calculated from the corresponding fourth period reaction number 214, and a fifth period reaction number 615 calculated from the corresponding fifth period reaction number 215 are provided.

  FIG. 11 is a diagram showing an example of the data structure of the document classification DB 7 and shows a portion of the category “Japanese movie” and the subcategory “1”.

  The document classification DB 7 has a common value for the category information 204 of the reaction history DB 2 and a record for each group when there are a plurality of one or more documents (referred to as groups) that are also common for the subcategories.

  The record includes the same category information 701 as the corresponding category information 204, the subcategory information 702 indicating the corresponding subcategory, the topic ratio 711 that is the ratio of words related to a predetermined topic (first topic) in the corresponding document, Topic ratio 712, which is the ratio of words related to another topic (second topic), topic ratio 713, which is the ratio of words related to another topic (third topic) in the document, and another topic ( A topic ratio 714 that is a ratio of words related to the fourth topic).

  FIG. 12 is a diagram showing an example of the data configuration of the topic DB 8 and shows a portion of the category “Japanese movie”.

  The topic DB 8 has a record for each group when there are a plurality of combinations (referred to as groups) of categories and topics.

  The record includes category information 801 indicating the corresponding category, topic number 802 indicating the corresponding topic, word information 811 indicating the first, second, and third words related to the topic in the corresponding document, word information 812, Word information 813 is provided.

  The number of word information (word information 811 and the like) may be 3 or less, 3 at the maximum, and 4 topics.

  Such a topic DB 8 has four topics, and applies LDA (Latend Dirichlet Allocation), LSA (Latent Semantic Analysis), pLSA (Probabilistic Latent Semantic Analysis), etc. to the part of the reaction history DB 2 shown in FIG. Can be generated.

  FIG. 13 is a flowchart showing a flow of processing for generating the document classification DB 7 by the reaction prediction acquisition unit 4.

  Here, a method for generating a record including category information 701 matching the target category for one category (hereinafter referred to as a target category) will be described. A record including category information 701 that matches another target category is generated in the same manner.

  The reaction prediction acquisition unit 4 searches the reaction history DB 2 for a record including the category information 204 that matches the target category, and extracts a past document 207 (hereinafter referred to as a past document) (S601).

  The reaction prediction acquisition unit 4 searches the topic DB 8 for a record including the category information 801 that matches the target category, and extracts the topic number 802 (S603).

  Here, it is assumed that four topic numbers (hereinafter referred to as topic numbers 8021, 8022, 8023, and 8024) are extracted.

  For each pair of past document and topic number, the reaction prediction acquisition unit 4 matches one of the word information 811, word information 812, and word information 813 in the record of the topic DB 8 that is in the past document and includes the topic number. The number of words to be calculated is obtained (S605).

  The reaction prediction acquisition unit 4 sums the number of words obtained for the topic number 8021 regardless of the past document (hereinafter, sum 80210), and sums of the number of words obtained for the topic number 8022 regardless of the past document (hereinafter, sum 80220). Then, the sum of the number of words obtained for the topic number 8023 regardless of the past document (hereinafter, sum 80230) is obtained (S607).

  For each past document, the response prediction acquisition unit 4 occupies a ratio (hereinafter referred to as a topic ratio 901) of the total number of words 80210 calculated for the past document and the topic number 8021, and a total number of words 80220 calculated for the topic number 8022. Occupying ratio (hereinafter referred to as topic ratio 902), ratio occupying the total number of words calculated for topic number 8023 (hereinafter referred to as topic ratio 903), and ratio occupying the total number of words calculated regarding topic number 8024 (hereinafter referred to as topic ratio 902). , A topic ratio 904) is generated (S609).

  FIG. 14 is a diagram illustrating an example of a past document vector of a document corresponding to each record of the reaction history DB 2 illustrated in FIG.

  For the past document with the document number “1”, the topic ratio 901 of the past document vector is 0.7, the topic ratio 902 is 0, the topic ratio 903 is 0, and the topic ratio 904 is 0.3. .

  For the past document with the document number “2”, the topic ratio 901 of the past document vector is 0.1, the topic ratio 902 is 0.5, the topic ratio 903 is 0, and the topic ratio 904 is 0.4. It is.

  For the past document with the document number “3”, the topic ratio 901 of the past document vector is 0.1, the topic ratio 902 is 0, the topic ratio 903 is 0.5, and the topic ratio 904 is 0.4. It is.

  For the past document with the document number “4”, the topic ratio 901 of the past document vector is 0.6, the topic ratio 902 is 0, the topic ratio 903 is 0.1, and the topic ratio 904 is 0.3. It is.

Returning to FIG. 13, the reaction prediction acquisition unit 4 calculates the distance d between the past document vectors for each set of the two past document vectors i and j by the following formula (S611).

  In the above example, since the number of elements is 4, zm is calculated as 4.

  Based on the distance d, the reaction prediction acquisition unit 4 uses a predetermined classification technique (for example, k-means clustering) to give subcategory information to each past document, and two past data to which the same subcategory information is given. The distance d for the document is made shorter than the distance d for the two past documents to which different subcategory information is assigned (S613).

  For example, the past documents with document numbers “1” and “4” shown in FIG. 14 are assigned a subcategory (subcategory information “1”), and the past document with document number “2” is assigned a subcategory (subcategory information “2”). )) And a past category with document number “3” is given a subcategory (subcategory information “3”).

  That is, four past documents are clustered into three.

  For each subcategory information, the reaction prediction acquisition unit 4 obtains the average value of the topic ratio 901, the average value of the topic ratio 902, the average value of the topic ratio 903, and the average value of the topic ratio 904 for the past document to which the subcategory information is assigned. Is calculated (S615).

  The reaction prediction acquisition unit 4 searches the document classification DB 7 for each subcategory information for a record including the subcategory information 702 that matches the subcategory information, and sets the average value of the topic ratio 901 as the average value of the topic ratio 711 and the topic ratio 902. Are stored as the topic ratio 713, the average value of the topic ratio 903 as the topic ratio 713, and the average value of the topic ratio 904 as the topic ratio 714 (S617), and the process ends.

  Since the processing of the document input unit 1 is the same as that of the first embodiment, description thereof is omitted.

  FIG. 15 is an overall flowchart showing a processing flow of the reaction prediction acquisition unit 4 and the reaction prediction output unit 5.

  The reaction prediction acquisition unit 4 decomposes the target document into words (S201), deletes unnecessary words (S203), detects the number of characters in the remaining word group (S204), and based on the word group and the target document category A subcategory of the target document (hereinafter referred to as a target document subcategory) is detected (S205), and a predicted reaction value is acquired (S208).

  Next, the reaction prediction output unit 5 displays the reaction prediction value on the reaction prediction screen 200 (S209), and the control returns to S101.

  FIG. 16 is a flowchart of target document subcategory detection (S205) by the reaction prediction acquisition unit 4.

  The reaction prediction acquisition unit 4 searches the topic DB 8 for a record including category information 801 that matches the target document category (S2051).

  The reaction prediction acquisition unit 4 obtains the number of words that are in the word group and match any one of the word information 811, 812, and 813 for each searched record, for example, for each record shown in FIG. 12 (S 2053). .

  The reaction prediction acquisition unit 4 generates a feature vector whose element is a ratio of the total number of words (total number of words) (S2055). For example, the reaction prediction acquisition unit 4 generates a feature vector whose elements are ratios corresponding to the subcategory information “1”, “2”, “3”, and “4” in FIG. 12, for example.

  The reaction prediction acquisition unit 4 searches the document classification DB 7 for records including the category information 701 that matches the target document category (S2057).

  The reaction prediction acquisition unit 4 generates a representative point vector having the topic ratio 711, the topic ratio 712, the topic ratio 713, and the topic ratio 714 as elements for each retrieved record (S2059).

The reaction prediction acquisition unit 4 calculates the distance D between the feature vector m and the representative point vector n by the following formula for each set of the feature vector m and the representative point vector n (S2061).

  In the above example, since the number of elements is 4, zm is calculated as 4.

  The reaction prediction acquisition unit 4 extracts the subcategory information 702, that is, the target document subcategory from the record of the document classification DB 7 corresponding to the minimum distance D (S2063), and ends the process.

  FIG. 17 is a flowchart of the reaction prediction value acquisition (S208) by the reaction prediction acquisition unit 4.

  The reaction prediction acquisition unit 4 determines from the reaction prediction DB 6 that (condition 1) the difference between the time 601 and the current time is within a predetermined range, (condition 2) whether the current day is a holiday or not matches the holiday flag 602. Search for records satisfying (Condition 3) The number of characters of the target document matches the number of characters 605, (Condition 4) The target document category and category information 603 match, and the target document subcategory and subcategory information 604 match. (S2081).

  The reaction prediction acquisition unit 4 sets the first period reaction number 611 as the first period reaction prediction value, the second period reaction number 612 as the second period reaction prediction value, the third period reaction number 613 as the third period reaction prediction value, The 4-period response number 614 is set as the fourth-period response predicted value, and the fifth-period response number 615 is set as the fifth-period response predicted value (S2087).

  Thus, by providing reaction prediction DB6 beforehand and taking out the reaction prediction value according to the object document from reaction prediction DB6, a reaction prediction value can be obtained rapidly.

  The process of the reaction prediction output unit 5 is the same as that in the first embodiment.

  For example, the target document is “Please tell me the title of the most interesting Japanese movie you have seen in the past. I really like action and horror. I would be happy if you could tell me in the movie. "

  The target document is decomposed into words (FIG. 15: S201), unnecessary words are deleted (FIG. 15: S203), and “recommended”, “interesting”, “Japanese film”, “title”, “movie” are included in the remaining word group. Will be included.

  For example, among the above five words, the words corresponding to the set of category information 701 “Japanese movie” and subcategory information “1” in FIG. 12 are “recommended” and “interesting”.

  The word corresponding to the set of category information 701 “Japanese film” and subcategory information “2” is “name”. There is no word corresponding to the set of the category information 701 “Japanese movie” and the subcategory information “3”. The words corresponding to the set of category information 701 “Japanese movie” and subcategory information “4” are “movie” and “Japanese movie”.

  Therefore, the feature vector is {2/5, 1/5, 0, 2/5} = {0.40, 0.20, 0, 0.40}.

  Here, the calculation was made with two significant figures.

  For example, the distance D between the representative point vector corresponding to the category information 701 “1” shown in FIG. 11 and the feature vector {0.40, 0.20, 0, 0.40} is 0.339. The distance D between the representative point vector corresponding to the category information 701 “2” and the feature vector is 0.424. The distance D between the representative point vector corresponding to the category information 701 “3” and the feature vector is 0.616.

  Therefore, the representative point vector corresponding to the category information 701 “1” is selected, that is, the target document subcategory is “1”.

  Further, since the number of characters of the target document is 99, the number of characters becomes 100 when the last digit is rounded off.

  In FIG. 10, assuming that the difference between the time 601 and the current time is within a predetermined range, and the status of whether or not the current day is a holiday is the same as the holiday flag 602, the uppermost record is Since the number of characters “100” matches the number of characters 605, the target document category and category information 603 match, and the target document subcategory and subcategory information 604 match, the first period response number 611 “6” (first period response prediction value) ), Second period response number 612 “4” (second period response predicted value), third period response number 613 “1” (third period response predicted value), fourth period response number 614 “1” (fourth) (Period response predicted value), 5th period response number 615 “0” (5th period response predicted value) is obtained.

  Thereby, for example, a reaction predicted value as shown in FIG. 3 is displayed.

  According to the second embodiment, the target document subcategory, that is, the category inherent in the target document is obtained, and the reaction prediction value is acquired based on this, so that the accuracy of the reaction prediction can be improved.

  By the way, in the social networking service (SNS), different communities are established for each user, and the behavior may be different. In this case, it is assumed that an appropriate response prediction value cannot be obtained for the new user's document.

  Therefore, the predicted response value is classified for each user who views the document, and the predicted response value for the new user's document is calculated based on the response that the browsing user has shown to the creator of the document in the past. A method is also conceivable in which, after the calculation is completed for the user who browses, the sum of the predicted reaction values is calculated and displayed on the screen.

  As described above, according to the document creation support apparatus according to the present embodiment, the response prediction acquisition for obtaining the feature amount of the target document and obtaining the reaction predicted value that is the predicted value of the response to the target document based on the feature amount is obtained. Therefore, the creator of the target document can know in advance the reaction to the target document based on the predicted response value, that is, can assist in creating a better target document.

  Further, by using the past history, for example, it is possible to obtain a predicted response value in consideration of the nature of the community that is the object of the document. For example, when a document is posted to a community with many housewives, it can be expected to obtain the knowledge that the most reaction is obtained at noon on weekdays and the reaction is less on holidays. In addition, the accuracy of estimation of reaction prediction values is improved by performing document classification using the above-described method using topics (latent topic estimation method). In addition, since the estimated value of the reaction of the responder (reaction predicted value) is displayed before the document is posted, the psychological load at the time of document posting is reduced.

  The document creation support apparatus has a plurality of records for a document that is not a target document, and each record includes a database (reaction history DB2 or reaction prediction DB6) that includes the feature amount of the corresponding document and the number of reactions to the document. ), And the reaction prediction acquisition unit searches the record corresponding to the feature quantity of the target document from the database and obtains the predicted response value based on the number of reactions included in the record, thereby supporting the creation of the target document. can do.

  In addition, since the feature amount includes the number of characters of the target document, creation of the target document can be supported by a predicted response value obtained based on the number of characters.

  Further, since the feature amount includes the number of characters of the target document, the category of the target document, and the subcategory of the target document, it can be expected to improve the accuracy of the predicted response value.

  In addition, a record for each group including a plurality of documents having a common category and subcategory is included, and the record includes a document classification DB including elements indicating characteristics of the group. Generate a feature vector indicating the feature of the document, search a record corresponding to the target document category from the document classification database, generate a representative point vector composed of elements included in the record for each searched record, Since the subcategory corresponding to the minimum distance is determined as the subcategory of the target document, the accuracy of the reaction predicted value can be expected to be improved.

  A computer program for causing a computer to function as a document creation support apparatus can be recorded on a computer-readable recording medium such as a semiconductor memory, a magnetic disk, an optical disk, a magneto-optical disk, a magnetic tape, or a communication network such as the Internet And can be widely distributed.

1 ... Document input part 2 ... Reaction history DB
3 ... Reaction acquisition unit 4 ... Reaction prediction acquisition unit 5 ... Reaction prediction output unit 6 ... Reaction prediction DB
7 ... Document classification DB
8 ... Topic DB
DESCRIPTION OF SYMBOLS 100 ... Document input screen 200 ... Reaction prediction screen 201 ... Document number 202, 601 ... Time 203, 602 ... Holiday flag 204, 603, 701, 801 ... Category information 205 ... Title 206, 605 ... Number of characters 207 ... Past document 211, 611 ... Number of first period reactions 212, 612 ... Number of second period reactions 213, 613 ... Number of third period reactions 214, 614 ... Number of fourth period reactions 215, 615 ... Number of fifth period reactions 300 ... Post button 604, 702 ... Subcategory information 711, 712, 713, 714, 911, 912, 913, 914 ... Topic ratio 802, 8021, 8022, 8023, 8024 ... Topic number 811, 812, 813 ... Word information 80210, 80220, 80230 ... Sum d, D ... distance i, j ... past document vector m ... features Vector n ... representative point vector

Claims (10)

A document creation comprising: a reaction prediction acquisition unit that obtains a feature amount of a target document that is a document creation support target and obtains a reaction prediction value that is a predicted value of a response to the target document based on the feature amount Support device. The document that is not the target document has a plurality of records, and each record includes a database that includes the feature amount of the corresponding document and the number of reactions to the document,
The reaction prediction acquisition unit
The document creation support apparatus according to claim 1, wherein a record corresponding to the feature amount of the target document is searched from the database, and the predicted response value is obtained based on the number of reactions included in the record. The document creation support apparatus according to claim 1, wherein the feature amount includes a number of characters after unnecessary characters are deleted from the target document. The document creation support apparatus according to claim 1, wherein the feature amount includes a number of characters after unnecessary characters are deleted from the target document, a category of the target document, and a subcategory of the target document. A record for each group including a plurality of documents having a common category and subcategory, and the record includes a document classification database including elements indicating characteristics of the group;
The reaction prediction acquisition unit
Generating a feature vector indicating features of the target document;
A record corresponding to the target document category is searched from the document classification database, and a representative point vector including elements included in the record is generated for each searched record.
Obtaining a distance between the feature vector and each representative point vector;
The document creation support apparatus according to claim 4, wherein a subcategory corresponding to a minimum distance is a subcategory of the target document. A document input part for inputting the target document;
A reaction prediction acquisition unit that obtains a feature amount of the target document and obtains a reaction prediction value that is a predicted value of a response to the target document based on the feature amount;
A document creation support apparatus comprising: a reaction prediction output unit that outputs the reaction prediction value. A step of obtaining a feature amount of the target document by the reaction prediction acquisition unit of the document creation support apparatus;
The operation method of the document creation support apparatus, characterized in that the reaction prediction acquisition unit includes a step of obtaining a reaction predicted value that is a predicted value of a response to the target document based on the feature amount. A document input unit of the document creation support apparatus inputting a target document that is a document creation support target;
A step of obtaining a feature amount of the target document by a reaction prediction acquisition unit of the document creation support apparatus;
The reaction prediction acquisition unit obtains a reaction predicted value that is a predicted value of a response to the target document based on the feature amount;
A response prediction output unit of the document creation support apparatus includes a step of outputting the reaction predicted value.   A computer program for causing a computer to function as the document creation support apparatus according to claim 1.   A computer program for causing a computer to function as the document creation support apparatus according to claim 6.

Images