JP5332918B2 - Classification data recommendation method, program, and apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To automatically check section data set to new text information, and to provide an inputting person with candidate information when the section data are erroneously added or not yet inputted. <P>SOLUTION: A classification code table generating part 101 generates a classification code table file 104 indicating a correlation between section data and classification codes, and a classification code statistical information table file 105 collecting the status of appearance in the population text file 109 of the classification codes from the population text file 109 of text information and the sample text file 110 of the text information to which section data determined to be right are added. A section data recommending part 102 extracts the classification codes from new text information, extracts the section data corresponding to the classification codes from the classification code table file 104, extracts statistical information from the classification code statistical information table file 105 for each section data, and then selects and presents candidates of the section data corresponding to the new text information on the basis of the statistical information. <P>COPYRIGHT: (C)2011,JPO&amp;INPIT

Description

  The present invention relates to a technique for accumulating text information together with segment data indicating the segment of the text information by one or more segment values.

  In recent years, there has been an increase in the number of companies that are focusing on call center information and conducting analysis. The background is the introduction of customer service systems and text mining software, and the establishment of the IT infrastructure of the call center.

  However, erroneous input tends to occur in the accumulated information for various reasons. For example, since there are many cases where a plurality of cases are input at a time after data input or after a call, the input person is often unaware of erroneous input.

Furthermore, the call center and customer service staff are often made up of part-time and temporary staff, and it is difficult to keep the input level constant due to regular personnel changes.
The following are examples of erroneous input.
(A) Typing error in text information (information entered directly) (B) Mistake in classification data (information selected from pull-down menu) (C) Classification data not set In addition, shown as (A) to (C) above In this example, it is assumed that the customer reception system 1801 as shown in FIG. The customer service system is a system that has a service history database 1802 in which an input person inputs a screen each time a service start time, customer information, question content, and service content, and manages the history.

The input screen of such a customer service system is often composed of four types of information, although there are differences depending on the type of business and business.

(1) Basic information A part for inputting information related to the reception of the project number, reception date and time, the receptionist, etc. (2) Customer information A part for inputting information about the customer such as a customer ID and a customer name (3) Inquiry information (4) Reply information The part to input the information about the answer contents and the answer category

In the above (1) to (4), there are items that are directly typed and items that are selected from pull-down menus, check boxes, and the like. In addition, an initial value may be set when the screen is started, or a database may be searched from a specific input value and another item value may be automatically set.

  An erroneous input is likely to occur when the question (3) and the answer (4) are entered. Actually, the typographical error in (A) often occurs in a part of the input characters. Since there are few cases where it does not make sense even if it occurs, there is little need to check and correct typographical errors.

  On the other hand, since the classification data of (B) and (C) is originally very troublesome to check the text content by visual inspection, the purpose of quickly grasping the tendency is to aggregate the classification data added by the input person. is there. Therefore, accuracy is required for the segment data.

  If the accuracy of the classification data is low, the validity of the tabulation results is lost. For this reason, some companies check the input data by visual inspection and correct the classification data. However, in many companies, there are many cases where there is a shortage of personnel and a large amount of accumulated information, and thus it is necessary to accumulate the data in the input stage. The classification data (B) and (C) are more serious than the problem (A) because the erroneous input continues when the input person does not know or misunderstands the input method.

JP 2001-060199 A JP 2002-183194 A JP 2002-189754 A

However, there has been no effective conventional technique for checking the erroneous input of the classification data as described above or supporting the input of the classification data.
Therefore, in one aspect of the present invention, it is possible to automatically check the classification data set on the input screen for new text information in the customer service system, etc., and when there is a possibility of a mistake or when it is not input, The purpose is to provide candidate information (recommendations).

In one example, the present invention is realized as a method of storing text information together with segment data indicating a segment of the text information by one or more segment values, and has the following configuration.
First, an arbitrary combination of a text information population text file to which classification data is assigned by each input person and a text information sample text file to which classification data determined to be correct is given in a predetermined text unit. As a key, the classification code table file indicating the correspondence between the classification data and the classification code is created.

  Next, for each category data extracted from the category code table file, category code statistics collected as statistical information on the appearance status of the category code associated with the category data in the category code table file in the population text file An information table file is created.

  Next, a classification code is extracted from the newly input new text information, and a classification code table file is searched for each extracted classification code, whereby classification data corresponding to the classification code is extracted.

  Then, statistical information is extracted from the classification code statistical information table file for each of the extracted classification data, and based on the extracted statistical information, the classification data candidates corresponding to the new text information are extracted from the classification data. Selected and presented.

  For the input of new text information, it becomes possible to accurately present the candidate (recommendation) of the classification data most suitable for it.

It is a block diagram of embodiment of the customer reception system which has the recommendation function of division data. It is an operation | movement flowchart which shows the classification code table preparation process which the classification code table preparation part 101 performs. It is an operation | movement flowchart which shows the control operation | movement of the division data recommendation display control process which the division data recommendation part 102 performs. 5 is an operation flowchart showing a control operation of classification code automatic recreation processing executed by a classification code automatic recreation unit 103; It is a figure which shows the example of a data structure of the population text file 109. FIG. It is a figure which shows the data structural example of the sample text file 110. FIG. It is a figure which shows the correspondence of classification data and classification hierarchy and classification code. 6 is a diagram showing an example data structure of a classification code table file 104. FIG. It is a figure which shows the data structural example of the classification code statistical information table file 105. FIG. It is a figure which shows the example of input of reception information, and the example of the morpheme, morpheme pattern, and morpheme matrix corresponding to it It is a figure which shows the example of a data structure of the morpheme matrix file which comprises a morpheme matrix. 6 is a diagram showing an example data structure of a search result file 107. FIG. It is a figure which shows the data structural example of the recommendation file 108. FIG. It is explanatory drawing of the display process of a recommendation message. It is a figure which shows the example of a display of the recommendation display of division data. It is a figure which shows the data structural example of the classification code statistical information table file 105 in the classification code automatic re-creation process. It is a figure which shows an example of the hardware constitutions of the computer which can implement | achieve embodiment of a customer service system. It is explanatory drawing of a customer reception system.

Hereinafter, embodiments will be described in detail.
FIG. 1 is a configuration diagram of an embodiment of a customer service system having a recommendation function for segment data.

  The customer service system 100 according to the present embodiment includes a classification code table creation unit 101, a classification data recommendation unit 102, and a classification code automatic re-creation unit 103. The classification code table creation unit 101 creates a classification code table file 104 and a classification code statistical information table file 105 based on a creation request from an input person. The classification data recommendation unit 102 searches the classification code table file 104 and generates a search result file 107 and a recommendation file 108 when a screen is input to the response history database 106 in customer response by an input person. Then, based on the contents of these files, the segment data recommendation unit 102 recommends the segment data, that is, presents candidates.

The classification code automatic re-creation unit 103 executes automatic re-creation of the classification code table file 104.
The operation of the embodiment of the customer reception system 100 having the above configuration will be described in detail below.

FIG. 2 is an operation flowchart showing a classification code table creation process executed by the classification code table creation unit 101 of FIG.
First, the input person designates the population text file 109 of the information source A and the sample text file 110 of the information source B. FIG. 5 is a diagram showing a data configuration example of the population text file 109, and FIG. 6 is a diagram showing a data configuration example of the sample text file 110.

  As shown in FIG. 5, the population text file 109 includes text information in which one record (data set) is input in response to a customer for a certain period in the past, and category 1 set by the input person for the text information. To segment N (for example, N = 3). Since this classification data is set by the judgment of the input person, an appropriate classification is not always set for the text information. The data amount of the population text file 109 is a large amount (for example, 10,000) because it is the total number of data obtained by customer correspondence in the past certain period (for example, one month). In this case, in order to synchronize with each file to be described later, a value obtained by combining the partition values from partition 1 to partition N constituting the partition data is used as the partition data ID.

  On the other hand, as shown in FIG. 6, in the sample text file 110, one record (data set) is composed of text information and segment data set optimally for the text information. The text information of the sample text file 110 and the classification data corresponding thereto are corrected so as to have an optimum correspondence relationship manually. The data amount of the sample text file 110 is, for example, about 50 records for one set of division data. As in the case of the population text file 109, a value obtained by combining the segment values from segment 1 to segment N constituting the segment data is used as the segment data ID.

  Each of the population text file 109 and the sample text file 110 may be a text file such as a CSV file, or may be a record file on a database system.

  The classification code table creation unit 101 in FIG. 1 creates the classification code table file 104 from the population text file 109 as the information source A and the sample text file 110 as the information source B as described below.

First, the classification code table creation unit 101 extracts classification codes from the population text file 109 and the sample text file 110 (step S201 in FIG. 2).
In this embodiment, in order to create the above-described classification code, the technique described in Japanese Patent Application No. 2008-258776 by the applicant of the present application is used. The outline of the processing by this technique is as follows.

Step 1
Morphological analysis is performed on the text information (see FIG. 6) of each record constituting the sample text file 110 in which the classification data is appropriately set, and a set of two morphemes commonly included in each record (two morphemes). Set) is extracted.

Step 2
Morphological analysis is also performed on the text information (see FIG. 5) of each record constituting the population text file 109, and the morpheme group constituting each text file 109 is extracted.
The classification hierarchy indicating the number of repetitions of extraction is set to 1.
The entire population text file 109 is selected as a set of population texts to be processed in the initial state.

Step 3
For each of the two morpheme sets extracted in step 1, the number of simultaneous occurrences in each record in the sample text file 110 (number of occurrences in the sample) and the number of simultaneous occurrences in the set of population text to be processed (mother) The number of occurrences in the group) is calculated. Then, “appearance rate = number of occurrences in sample ÷ number of occurrences in population” is calculated.
A group of two morpheme sets whose appearance rate is within the upper predetermined order is selected as a classification code group in the current classification hierarchy (initially 1).

Step 4
The classification data set in the sample text file 110 including each classification code is associated with each classification code (= 2 morpheme set) selected in step 3 and the data set corresponding to the current classification hierarchy. .

Step 5
The set of text information in the population text file 109 that includes only each of the classification codes (= 2 morpheme pairs) selected in step 3 is set as a new set of population texts to be processed. , Steps 3 and 4 and Step 5 are repeatedly executed.
In step 5, when the change in the number of text information included in the set of population texts is equal to or less than a predetermined threshold, the classification code extraction process is terminated.

  FIG. 7 is a diagram showing a correspondence relationship between the classification data generated by the repetitive processing from step 1 to step 5 and the classification hierarchy / classification code. The classification code is a candidate for a set of two appropriate morphemes (two morpheme sets) to be included in the text file as a result of morphological analysis for the text file when certain classification data is set for the text file. It is. A plurality of classification codes may be associated with one type of classification data. In addition, the same classification code with different classification hierarchy may be associated with one type of classification data multiple times.

  As shown in step 5 above, in the extraction of the classification code, the population text file is narrowed down as the classification hierarchy progresses. For this reason, when the morpheme obtained by morphological analysis of a newly input text file includes two morphemes constituting a classification code selected over a plurality of classification hierarchies, the following estimation is performed. Is possible. That is, it can be estimated that the newly input text file has more similar text content to the sample text file 110 in which the classification data associated with the classification code is set. For this reason, it is recommended to set the classification data in the newly input text file. This is the basic concept of the partitioned data recommendation function in this embodiment.

  Next, the classification code table creation unit 101 in FIG. 1 creates the classification code table file 104 based on the above-described classification code extracted in step S201 in FIG. 2 (step S202 in FIG. 2). FIG. 8 is a diagram showing a data configuration example of the classification code table file 104 generated based on the correspondence relationship of FIG. “Classification lowest layer” data is added to each format (record) in the correspondence format of FIG. This data is information indicating to which classification hierarchy the corresponding classification code has been extracted. For example, the classification code “delivery form-print” of the first record shown in FIG. 8 is data of the classification hierarchy 1, but this classification code is also selected in the classification hierarchy 3. This information is used when the classification data recommendation unit 102 in FIG. 1 executes a classification code table search process described later.

  The classification code table file 104 may be a text file such as a CSV file or a record file on a database system. One row in FIG. 8 represents one data set, that is, one record. A value obtained by combining the segment values from segment 1 to segment N constituting the segment data is used as the segment data ID.

  Next, the classification code table creation unit 101 creates the classification code statistical information table file 105 based on the content of the classification code extraction process in step S201 in FIG. 2 (step S203 in FIG. 2).

  FIG. 9 is a diagram illustrating a data configuration example of the classification code statistical information table file 105. This file consists of category data items (sets of category values from category 1 to category N), classification code statistical information items of R1 to R4 and R6 to R8, classification accuracy items, and creation date items. Composed. A value obtained by combining the segment values from the segment 1 to the segment N of the segment data item is used as the segment data ID.

  Based on the processing result in the operation flowchart of FIG. 2, the information of the record group having the same classification data among the classification data registered in the classification data item of the classification code table file 104 is the classification code statistical information table file. One record 105 is collected. And the item of each processing object record of classification code statistics information table file 105 is generated as follows.

In the R1 item, the number of records of the population text file 109 that is the basis of the classification code table file 104 is registered.
In the R2 item, the number of records of the sample text file 110 is registered.

  In the R3 item, the number of records extracted when the population text file 109 is searched using one or more classification codes associated with the classification data is registered. More specifically, the classification code table creation unit 101 searches the classification data items in the classification code table file 104 of FIG. 8 with the classification data of the processing target records in the classification code statistical information table file 105. Then, the classification code table creation unit 101 extracts the classification code stored in the classification code item (see FIG. 8) of each record extracted by the search. Next, the classification code table creation unit 101 searches the text information item (see FIG. 5) of the population text file 109 for each extracted classification code, and records the records that include both of the two morphemes constituting the classification code. Extract the number of records. Then, the classification code table creation unit 101 adds the number of retrieved records for the entire extracted classification code, and registers the result in the R3 item.

  In the R4 item, the number of records in which the same classification data as the classification data of the processing target record in the classification code statistical information table file 105 is set among the records retrieved from the population text file 109 in the determination of the R3 item Is registered.

In the classification accuracy item, a value obtained by dividing the item value of R4 by the item value of R3 is registered.
The creation date is the date of the record including the text information (FIG. 5) registered most recently in the population text file 109.

R6 to R8 are used by the automatic classification code re-creation unit 103 in FIG. 1, and the initial values are all zero. Details of these will be described later.
As described above, when the classification code table creation unit 101 in FIG. 1 creates the classification code table file 104 and the classification code statistical information table file 105, the classification data recommendation unit 102 in FIG. 1 becomes operable.

FIG. 3 is an operation flowchart illustrating the control operation of the partition data recommendation display control process executed by the partition data recommendation unit 102.
First, the input person (see FIG. 1) inputs the content of the customer response using the screen input unit of the customer response system 100 (step S301 in FIG. 3). FIG. 10A is a diagram illustrating an input example. Along with the text of the question content, classification data (larger classification, middle classification, etc. can be further classified) is input.

  In response to this, the classification data recommendation unit 102 in FIG. 1 first issues one search ID for the input content in step S301 (step S302 in FIG. 3). The search ID is, for example, “A00000000000000”.

  The above input contents are sequentially stored together with the search ID and creation date in the response history database 106 of FIG.

  Next, the classification data recommendation unit 102 extracts text data of the question content from the input content in step S301, and executes a morphological analysis process on the data (step S303 in FIG. 3). As a result, for example, morphemes as illustrated in FIG. 10B are extracted from the text data of the question content portion in the input example of FIG.

  Next, the partitioned data recommendation unit 102 selects a morpheme pattern composed of two morphemes from the morphemes extracted in step S303 as illustrated in FIG. 10C, and creates a morpheme matrix file based thereon. (Step S304 in FIG. 3). FIG. 10D is a diagram conceptually showing a morpheme matrix, in which morphemes that can appear in columns and rows are arranged, and the intersection of each column and each row consists of the morpheme of each column and the morpheme of each row. Represents a morpheme pattern. The part of x corresponding to the pair of the same morpheme and the pair of morpheme whose order is reversed is excluded, and only the morpheme pattern of the part of ○ is extracted as data constituting the element of the morpheme matrix. FIG. 11 is a diagram illustrating a data configuration example of a morpheme matrix file that forms a morpheme matrix. For each search ID issued in step S302, that is, for each input content in step S301, a morpheme pattern constituting an element of the morpheme matrix is registered.

Next, the segment data recommendation unit 102 repeatedly executes the following series of processes by the loop process from step S305 to S310.
First, the partitioned data recommendation unit 102 sequentially selects morpheme patterns that are elements of the morpheme matrix from the morpheme matrix file obtained in step S304 (step S306 in FIG. 3).

  Next, the classification data recommendation unit 102 extracts a record having the morpheme pattern selected in step S306 as the value of the classification code item in the classification code table file 104 (see FIG. 8). Subsequently, the classification data recommendation unit 102 determines the contents of the classification hierarchy item and the classification bottom layer item of the extracted record. Based on this determination, the classification data recommendation unit 102 searches the classification data (= classification ID) from which the records of all hierarchies from the first hierarchy to the hierarchy indicated by the classification lowest layer item are extracted (see FIG. 3 step S307).

  Next, the segment data recommendation unit 102 determines whether or not the search in Step S307 has been hit (Step S308 in FIG. 3). Then, when a hit is found, the category data recommendation unit 102 registers the hit category data in the search result file 107 in association with the search ID issued in step S302 (step S309 in FIG. 3). . FIG. 12 is a diagram illustrating a data configuration example of the search result file 107. In each record of the search result file 107, the search ID issued in step S302 is stored in the search ID item, and the category data hit in the search in step S307 is stored in the category data item.

The series of processing from step S306 to S309 is repeatedly executed for each morpheme pattern constituting the elements of the morpheme matrix obtained from the input contents of the customer service.
When the iterative processing from steps S305 to S310 for all morpheme patterns in the morpheme matrix is completed, the partitioned data recommendation unit 102 performs the next processing target record newly registered in the search result file 107 in step S309. Execute the process. That is, the category data recommendation unit 102 searches the category code statistical information table file 105 for a record whose content of the category data item is the same as the category data of the record to be processed (see FIG. 12). Next, the classification data recommendation unit 102 extracts the classification accuracy value stored in the classification accuracy item of the retrieved record. Then, the classification data recommendation unit 102 combines the extracted classification accuracy value and the contents of the processing target record. Thereafter, the classification data recommendation unit 102 registers the classification accuracy value up to the upper predetermined rank (for example, the third rank) and the contents of the processing target record combined therewith in the recommendation file 108 (step S311 in FIG. 3). ). FIG. 13 is a diagram illustrating a data configuration example of the recommendation file 108. Each record of the recommendation file 108 includes a search ID item and a classification data item in which the search ID and classification data (see FIG. 12) of the processing target record are stored, and a classification accuracy item in which a classification accuracy value is stored. The

  Next, the category data recommendation unit 102 performs a comparison process between the category data (see FIG. 10A) of the response contents being input in step S301 and the recommendation file 108 (step S312 in FIG. 3). Specifically, the classification data recommendation unit 102 registers, in the recommendation file 108 (see FIG. 13), a record whose search ID item value is the same as the search ID issued in step S302 is registered in the classification accuracy item. Extraction with high classification accuracy. Then, the category data recommendation unit 102 compares the category data in the response contents with the category data (see FIG. 13) in the order of the records extracted from the recommendation file 108. To do.

  Based on the result of the comparison determination at step S312, the segment data recommendation unit 102 displays a message indicating the recommendation of the segment data in correspondence with the contents of the response being input (step S313 in FIG. 3). FIG. 14 is a diagram illustrating the relationship between the comparison determination in step S312 and the control process for displaying a recommendation message executed in response to the comparison result. If it is determined that the category data in the response content matches the category data of the first rank record extracted from the recommendation file 108, that is, the classification accuracy, the recommendation message is displayed. Not. This means that the designation of the classification data by the input person is optimal. Also, if it is determined that the category data in the response content matches the category data of the records other than the first one extracted from the recommendation file 108, that is, the classification accuracy is the second or lower record, the match detection rank The classification data of the higher-order record is recommended. This means that there is optimum segment data than designation of segment data by the input person. Further, if it is determined that the category data in the contents of the response does not match the category data of any record extracted from the recommendation file 108, the category data of all the extracted rank records are recommendedly displayed. This means that it is not appropriate to specify the category data by the input person. In this case, the category data candidates of all ranks are recommended and displayed. FIG. 15 is a diagram illustrating a display example of the recommendation display of the segment data displayed to the input person by the above-described processing.

  In the display illustrated in FIG. 15, the input user can adopt the recommended category data by clicking the “select” link on the right side of the recommendation of any category data. The category data recommendation unit 102 accepts a recommended category data selection operation by the input person (step S314 in FIG. 3), and automatically sets the selected category data in the category data input form area in the response input screen ( Step S315 in FIG. 3).

Thereafter, the processing of the segment data recommendation unit 102 ends.
In the embodiment described above, it is possible to accurately present a candidate (recommendation) of classification data most suitable for input of new response content.

FIG. 4 is an operation flowchart showing the control operation of the automatic classification code re-creation process executed by the automatic classification code re-creation unit 103.
This process uses the classification code statistical information table file 105 to automatically analyze the degree of coincidence of the category data and automatically recreate the category code of the category data having a low degree of coincidence. By automating this process, it is not necessary to manually check the classification data, the classification code can be automatically learned, and the recommendation process can be performed with high accuracy.

In this process, first, the following check process is executed at regular intervals.
First, the classification code automatic re-creation unit 103 initializes each item value of R6, R7, and R8 in each classification data record of the classification code statistical information table file 105 having the data configuration example shown in FIG. A value 0 is set (step S401 in FIG. 4).

Next, the classification code automatic re-creation unit 103 sets the following calculation result in the R6 item in each classification data record of the classification code statistical information table file 105 (step S402 in FIG. 4). The R4 item value and the R1 item value are values in the same record as the R6 item.

R6 item value = R4 item value / R1 item value

As for the R6 item value calculated in this way, there is a record in the population text file 109 having the record number indicated by the R1 item value that matches the segment data having the record number indicated by the R4 item value. The ratio, that is, the expected value is shown.

  Next, the automatic classification code re-creation unit 103 takes out the combination of the classification data and creation date of each record from the classification code statistical information table file 105 (see FIG. 9) (step S403 in FIG. 4).

Next, the classification code automatic recreating unit 103 repeatedly executes the following series of processes for each combination of the classification data extracted in step S403 and the creation date (loop processing from steps S404 to S408 in FIG. 4).
That is, first, the classification code automatic re-creation unit 103 selects one combination of the classification data and the creation date (step S405 in FIG. 4).

  Next, the automatic classification code re-creation unit 103 searches the response history database 106 for the number of records in the response history in which the creation dates after the creation date of the set selected in step S405 are stored. Then, the classification code automatic re-creation unit 103 sets the number of records in the R7 item of the record corresponding to the group selected in step S405 in the classification code statistical information table file 105 (step S406 in FIG. 4). .

  Next, the classification code automatic re-creation unit 103 stores the creation date after the creation date of the group selected in step S405 from the response history database 106, and the same segment data as the segment data of the group is stored. Search the number of records in the response history. Then, the classification code automatic re-creating unit 103 sets the number of records in the R8 item of the record corresponding to the group selected in step S405 in the classification code statistical information table file 105 (step S407 in FIG. 4). .

  The series of processes in steps S405 to S407 described above is executed for all records on the classification code statistical information table file 105 (loop process from steps S404 to S408 in FIG. 4).

Thereafter, the classification code automatic re-creation unit 103 extracts classification data that satisfies the following conditions from the classification code statistical information table file 105 (step S409 in FIG. 4).

R8 item value / R7 item value (actually existing ratio) <R6 item value (expected ratio) × α

Here, α is a threshold system parameter and is a value that can be freely set by the administrator.

  FIG. 16 is a diagram illustrating a data configuration example of the classification code statistical information table file 105 in the automatic classification code re-creation process, and corresponds to the data configuration example of FIG. In the record shown in this example, the number of records R4 having the classification data of the record for the population text file 109 of R1 = 10,000 was originally 180. Therefore, R6 = R4 / R1 = 180 / 10,000 = 0.018. The record number R7 registered in the response history database 106 after the latest creation date of the record is 20,000. Among these 20,000 records, the record count R8 having the category data matching the category data of the record was 100. Accordingly, in step S409, the segment data of the records in which the actually existing ratio (R8 / R7 = 0.01) is smaller than the expected ratio (R6 = 0.018) × α is extracted.

  Finally, the classification code automatic recreating unit 103 executes the following processing when the classification data that matches the above-described condition is extracted in step S409. That is, the classification code automatic re-creation unit 103 sets the record group on the response history database 106 corresponding to R7 as the population text file 109 and the record group on the response history database 106 corresponding to R8 as the sample text file 110. Then, the classification code automatic re-creation unit 103 automatically regenerates the classification code table file 104 and the classification code statistical information table file 105 by executing the same classification code creation process as described in the operation flowchart of FIG. create.

FIG. 17 is a diagram illustrating an example of a hardware configuration of a computer that can realize the embodiment of the customer reception system described above.
The computer shown in FIG. 17 includes a CPU 1701, a memory 1702, an input device 1703, an output device 1704, an external storage device 1705, a portable recording medium driving device 1706 into which a portable recording medium 1709 is inserted, and a network connection device 1707. However, they are connected to each other by a bus 1708. The configuration shown in the figure is an example of a computer that can implement the above system, and such a computer is not limited to this configuration.

  The CPU 1701 controls the entire computer. The memory 1702 is a memory such as a RAM that temporarily stores a program or data stored in the external storage device 1705 (or the portable recording medium 1709) when executing a program, updating data, or the like. The CUP 1701 performs overall control by reading the program into the memory 1702 and executing it.

  The input device 1703 includes, for example, a keyboard, a mouse, etc. and their interface control devices. The input device 1703 detects an input operation by a user using a keyboard, a mouse, or the like, and notifies the CPU 1701 of the detection result.

  The output device 1704 includes a display device, a printing device, etc. and their interface control devices. The output device 1704 outputs data sent under the control of the CPU 1701 to a display device or a printing device.

The external storage device 1705 is, for example, a hard disk storage device. Mainly used for storing various data and programs.
The portable recording medium driving device 1706 accommodates a portable recording medium 1709 such as an optical disk, SDRAM, or compact flash, and has an auxiliary role for the external storage device 1705.

The network connection device 1707 is a device for connecting, for example, a LAN (local area network) or WAN (wide area network) communication line.
The system according to the embodiment is realized by the CPU 1701 executing a program equipped with the function of each block shown in FIG. 1 or the function corresponding to the processing of the operation flowcharts shown in FIGS. The program may be distributed by being recorded in, for example, the external storage device 1705 or the portable recording medium 1709, or may be acquired from the network by the network connection device 1707. Further, data used in each process is read from the external storage device 1705 to the memory 1702 and processed, for example.

  In the embodiment described above, the embodiment in which the recommendation display is performed when the classification data is given when the customer reception information is input in the customer reception system has been described. However, the above technique is a system that provides the classification data when inputting some text information. It can be widely applied to.

  It can be used to input segment data in customer service systems used in call centers and customer service rooms.

100, 1801 Customer response system 101 Classification code table creation unit 102 Classification data recommendation unit 103 Automatic classification code re-creation unit 104 Classification code table file 105 Classification code statistical information table file 106, 1802 Response history database 107 Search result file 108 Recommendation file 109 Population text file (source A)
110 Sample text file (source B)
1701 CPU
1702 Memory 1703 Input device 1704 Output device 1705 External storage device 1706 Portable recording medium driving device 1707 Network connection device 1708 Bus 1709 Portable recording medium

Claims (7)

In a method for a computer to store text information together with segment data indicating a segment of the text information by one or more segment values,
The computer
A population text file of text information to which the classification data is assigned by an input person, and a sample text file of text information to which the classification data determined to be correct are given in an arbitrary combination of predetermined text units. By creating a classification code table file indicating the correspondence between the classification data and the classification code by comparing the configured classification code as a key,
For each classification data extracted from the classification code table file, a classification in which the appearance status of the classification code associated with the classification data in the classification code table file in the population text file is collected as statistical information A classification corresponding to the classification code is created by creating a code statistical information table file, extracting the classification code from newly input new text information, and searching the classification code table file for each extracted classification code Extract the data,
Extracting the statistical information from the classification code statistical information table file for each of the extracted classification data;
Based on the extracted statistical information, a candidate for segment data corresponding to the new text information is selected from the extracted segment data and presented.
A partitioned data recommendation method characterized by that. Storing the new text information as a history database;
For each of the classification data extracted from the classification code table file, calculate the appearance status of the classification data in the history database,
Comparing the appearance status with the statistical information extracted from the classification code statistical information table file corresponding to the classification data;
Based on the comparison result, the population text file and the sample text file are determined from the history database, and the classification code table file and the classification code statistical information table file are recreated.
The partitioned data recommendation method according to claim 1. The text unit is a morpheme;
3. The partitioned data recommendation method according to claim 1 or 2, wherein: In a computer that stores text information together with segment data indicating a segment of the text information by one or more segment values,
A population text file of text information to which the classification data is assigned by an input person, and a sample text file of text information to which the classification data determined to be correct are given in an arbitrary combination of predetermined text units. By creating a classification code table file indicating the correspondence between the classification data and the classification code by comparing the configured classification code as a key,
For each classification data extracted from the classification code table file, a classification in which the appearance status of the classification code associated with the classification data in the classification code table file in the population text file is collected as statistical information A classification corresponding to the classification code is created by creating a code statistical information table file, extracting the classification code from newly input new text information, and searching the classification code table file for each extracted classification code Extract the data,
Extracting the statistical information from the classification code statistical information table file for each of the extracted classification data;
Based on the extracted statistical information, a candidate for segment data corresponding to the new text information is selected from the extracted segment data and presented.
Program for executing processing. Storing the new text information as a history database;
For each of the classification data extracted from the classification code table file, calculate the appearance status of the classification data in the history database,
Comparing the appearance status with the statistical information extracted from the classification code statistical information table file corresponding to the classification data;
Based on the comparison result, the population text file and the sample text file are determined from the history database, and the classification code table file and the classification code statistical information table file are recreated.
The program according to claim 1, further executing a process. In an apparatus for storing text information together with segment data indicating a segment of the text information by one or more segment values,
A population text file of text information to which the classification data is assigned by an input person, and a sample text file of text information to which the classification data determined to be correct are given in an arbitrary combination of predetermined text units. By comparing the configured classification code as a key, a classification code table file indicating a correspondence relationship between the classification data and the classification code is created, and for each of the classification data extracted from the classification code table file, A classification code table creation unit that creates a classification code statistical information table file that collects the appearance status of the classification code associated with the classification data in the classification code table file as statistical information;
The classification code is extracted from the newly input new text information, and the classification code table file is searched for each extracted classification code, so that the classification data corresponding to the classification code is extracted and the extracted The statistical information is extracted from the classification code statistical information table file for each classification data, and based on the extracted statistical information, a classification data candidate corresponding to the new text information is selected from the extracted classification data And a segment data recommendation part to be presented,
A segmented data recommendation device comprising: The new text information is accumulated as a history database, and for each category data extracted from the classification code table file, the appearance status of the category data in the history database is calculated, and the appearance status is used as the category data. Correspondingly, the statistical information extracted from the classification code statistical information table file is compared, and based on the comparison result, the population text file and the sample text file are determined from the history database, and the classification code table A classification code automatic re-creation unit that re-creates the file and the classification code statistical information table file;
The classified data recommendation device according to claim 1.