JP7555274B2 - Proposed device, proposed method and program - Google Patents

Description

The present invention relates to a proposal device, a proposal method, and a program.

In recent years, VOC (Voice of Customer) analysis, which analyzes a large amount of customer requests and evaluations (praise and complaints about a company's services), has become popular (see Patent Document 1). In VOC analysis, a large amount of customer requests and evaluations are collected by, for example, surveys such as online reviews on the Internet's SNS (Social Networking Service), questionnaire surveys conducted by the company, and surveys by call centers. In addition, various systems have been developed to analyze the collected requests and evaluations. The results of the analysis conducted by these various systems are fed back to improve the company's services and products.

The large amount of collected data on requests and evaluations may include evaluation points (quantitative data) that express evaluations in quantitative terms, such as a five-point scale, and text data (qualitative data) that express evaluations in free-form text using natural language. When analyzing text data, text mining techniques are often used. Furthermore, analyzing text data is technically more difficult than analyzing quantitative data, so analyzing text data often involves visualizing the text data.

JP 2014-191556 A

The results that the managers and staff of their company's services (hereinafter referred to as "Managers, etc.") really want are not the results of visualizing (analyzing) the large amount of text data, including requests and evaluations, obtained from customers, but measures (next actions) derived from the results of the visualization. Based on the measures derived, the managers, etc. of their company's services can improve customer satisfaction (evaluation points by customers) with their company's services.

However, in conventional technology, a large amount of customer requests and evaluations are analyzed by a person, and the analysis tends to be narrowed down based on the person's (the analyst's) subjective opinion. For example, from the perspective of time and low cost, there is a tendency for only comments (text data) associated with particularly low evaluation points on a five-point scale to be analyzed out of the large amount of evaluations obtained from customers. When the analysis target is narrowed down in this way, the accuracy of measures to improve or maintain customer satisfaction may decrease.

In view of the above circumstances, the present invention aims to provide a proposal device, a proposal method, and a program that can improve the accuracy of measures to improve or maintain customer satisfaction.

One aspect of the present invention is a proposal device that includes a control unit that acquires evaluation points and text data obtained from multiple customers, an analysis unit that executes a predetermined analysis process on the text data, an analysis unit that executes a weighting process on the results of the predetermined analysis process using the multiple evaluation points ranging from positive to negative, and a creation unit that creates data representing measures related to the evaluation points based on the results of the weighting process.

In one aspect of the present invention, in the proposed device described above, the analysis unit executes, as the predetermined analysis process, a process of classifying the text data, a process of extracting a plurality of elements from the classified text data, and a clustering process for the extracted plurality of elements.

In one aspect of the present invention, the proposed device as described above, the analysis unit performs a weighting process on the multiple elements, and the creation unit prioritizes the elements with higher weightings and creates data representing measures related to the evaluation points associated with the prioritized elements.

One aspect of the present invention is the proposed device described above, in which the control unit acquires the time-series evaluation scores and the time-series text data, the analysis unit executes the predetermined analysis process on the time-series text data, the analysis unit executes the time-series weighting process using the time-series evaluation scores, and the creation unit creates data representing measures related to the time-series evaluation scores based on the results of the time-series weighting process.

One aspect of the present invention is a proposal method executed by a proposal device, the proposal method including a control step of acquiring evaluation points and text data obtained from a plurality of customers, an analysis step of executing a predetermined analysis process on the text data, an analysis step of executing a weighting process on the results of the predetermined analysis process using a plurality of the evaluation points ranging from positive to negative, and a creation step of creating data representing measures related to the evaluation points based on the results of the weighting process.

One aspect of the present invention is a program for causing a computer to function as the proposed device described above.

This invention makes it possible to improve the accuracy of measures to improve or maintain customer satisfaction.

FIG. 1 is a diagram illustrating an example of a configuration of a proposed system in a first embodiment. FIG. 4 is a diagram showing an example of quantitative data and qualitative data in the first embodiment. FIG. 11 is a diagram showing an example of a classification result of qualitative data in the first embodiment. 11A and 11B are diagrams illustrating an example of an extraction result of elements of a classification result in the first embodiment. FIG. 11 is a diagram showing an example of a clustering result of extraction results in the first embodiment. FIG. 11 is a diagram showing a display example of a proposed content in the first embodiment. 4 is a flowchart showing an example of the operation of the proposed device in the first embodiment. FIG. 11 is a diagram showing an example of quantitative data and qualitative data in the second embodiment. FIG. 13 is a diagram illustrating an example of a time-series key performance indicator in the second embodiment. FIG. 11 is a diagram showing a display example of proposal content in the second embodiment. FIG. 2 is a diagram illustrating an example of a hardware configuration of a proposed device in each embodiment.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described in detail with reference to the drawings.
First Embodiment
FIG. 1 is a diagram showing a configuration example of a proposal system 1. The proposal system 1 proposes measures to improve customer satisfaction (evaluation points obtained from multiple customers) based on evaluation points (quantitative data) that express evaluation in quantitative expressions (numerical values) such as multi-level evaluation, and text data (qualitative data) that express evaluation in free description using natural language. Customer satisfaction may be a key performance indicator (KPI). In addition, the proposed measures may be measures to be continued for positive elements (elements that are well-received by customers), or may be new measures (points for improvement) to be implemented for negative elements (elements that are unpopular with customers).

The proposal system 1 includes a user terminal 2, a communication line 3, and a proposal device 4. The user terminal 2 includes an operation unit 20, a data processing unit 21, a communication unit 22, and a display unit 23. The proposal device 4 includes a communication unit 40, a control unit 41, a storage device 42, an analysis unit 43, an analysis unit 44, and a creation unit 45. The proposal system 1 may include multiple user terminals 2. The analysis unit 43 includes a classification unit 430, an extraction unit 431, and a clustering unit 432.

The user terminal 2 is an information processing device such as a personal computer, a tablet terminal, or a smartphone terminal. The user terminal 2 collects evaluation points (quantitative data) that express evaluations in numerical expressions and text data (qualitative data) that express evaluations in free descriptions using natural language. Such evaluation points and text data (text data that express evaluations) are collected for a predetermined theme as data on requests and evaluations obtained from customers by, for example, surveys such as word-of-mouth on SNS on the Internet, questionnaire surveys conducted by the company, and surveys by call centers. The user terminal 2 records evaluation points (quantitative data) that express evaluations in numerical expressions, text data (qualitative data) that express evaluations in free descriptions using natural language, and data that express the date on which the data was collected in the proposal device 4 for each predetermined theme. The text data may be text data generated by the customer operating a predetermined operation unit (for example, a keyboard, a touch panel, etc.) or text data generated by recognizing the customer's voice.

The operation unit 20 is, for example, an operation device such as a keyboard, a mouse, or a touch panel. The operation unit 20 accepts operations by a user. This user is not limited to a specific user, but may be, for example, a person in charge of the company's services or products.

The data processing unit 21 outputs to the communication unit 22, for each predetermined theme, evaluation scores (quantitative data) that express evaluation in quantitative terms, text data (qualitative data) that express evaluation in free description using natural language, and data that expresses dates. These data are transmitted by the communication unit 22 to the proposal device 4, and are recorded in the storage device 42.

The data processing unit 21 may generate parameters in response to a user operation received by the operation unit 20. The data processing unit 21 may output the generated parameters to the communication unit 22. The parameters output to the communication unit 22 are transmitted to the proposal device 4 by the communication unit 22. The parameters transmitted to the proposal device 4 are used as parameters to be used when the proposal device 4 executes an analysis process, an analysis process, etc.

The data processing unit 21 acquires the image data transmitted from the proposed device 4 from the communication unit 22. The data processing unit 21 generates an image based on the image data transmitted from the proposed device 4. The data processing unit 21 outputs the generated image to the display unit 23.

The communication unit 22 communicates with the communication unit 40 via the communication line 3. For example, the communication unit 22 transmits (uploads) to the communication unit 40 an evaluation score that expresses the evaluation in quantitative terms, text data that expresses the evaluation in free description using natural language, and data that expresses a date, for each predetermined theme. For example, the communication unit 22 may transmit an identification number of its own device to the communication unit 40.

The communication unit 22 acquires the data transmitted from the communication unit 40. The data transmitted from the communication unit 40 is, for example, data representing measures (insights) related to improving or maintaining customer satisfaction. The format of the data transmitted from the communication unit 40 is not limited to a specific format, but is, for example, an image data format. The communication unit 22 outputs the data transmitted from the communication unit 40 to the data processing unit 21.

The display unit 23 is a display device such as a liquid crystal display or an organic EL (Electro Luminescence) display. The display unit 23 displays an image based on the data transmitted from the proposal device 4. The data-based image is an image of a sentence (character string) representing a measure proposed by the proposal device 4. The data-based image may be an image of a sentence (character string) representing a customer comment (specific opinion) extracted from text data by the proposal device 4. The data-based image may also be a graph in a predetermined format, such as a graph with clustering (e.g., a scatter graph), a pie chart, or a bar graph. With these, the display unit 23 functions as a dashboard.

The communication line 3 is a communication line such as the Internet. The communication line 3 transmits data between the user terminal 2 and the proposed device 4, for example.

The proposal device 4 is an information processing device such as a server, a personal computer, a tablet terminal, or a smartphone terminal. The proposal device 4 provides a user with a service that proposes measures to improve customer satisfaction (key performance indicators). The service that proposes measures may be provided to the user as one of the cloud services.

The proposal device 4 performs an analysis process on text data (qualitative data) expressing evaluations in free-form textual descriptions using natural language, based on parameters transmitted from the user terminal 2. The proposal device 4 performs an analysis process on the results of the analysis process, based on evaluation points (weighting coefficients) expressing evaluations in quantitative terms. The proposal device 4 proposes measures to improve or maintain customer satisfaction to the user, based on the results of the analysis process.

The communication unit 40 communicates with the communication unit 22 via the communication line 3. For example, the communication unit 40 acquires from the communication unit 22 an evaluation score (quantitative data) that expresses the evaluation in quantitative terms, text data (qualitative data) that expresses the evaluation in free description using natural language, and data that expresses a date, for each specified theme. The communication unit 40 transmits the data to the communication unit 22. For example, the communication unit 40 transmits data that expresses measures for improving or maintaining customer satisfaction to the communication unit 22.

The control unit 41 controls the operation of each functional unit of the proposal device 4. The control unit 41 acquires, for each predetermined theme, an evaluation score expressing the evaluation in quantitative terms, text data expressing the evaluation in free description using natural language, and data expressing a date from the communication unit 40. The control unit 41 outputs the text data expressing the evaluation to the classification unit 430. The control unit 41 acquires, from the clustering unit 432, the results of the clustering process performed by the clustering unit 432.

The control unit 41 outputs the evaluation score (quantitative data) representing the evaluation and the results of the clustering process to the analysis unit 44. The control unit 41 acquires data representing measures for improving or maintaining customer satisfaction from the creation unit 45. The control unit 41 outputs the data representing measures for improving or maintaining customer satisfaction to the communication unit 40.

The control unit 41 may record data representing measures for improving or maintaining customer satisfaction in the storage device 42. This allows the user terminal 2 to download data representing past measures from the proposal device 4.

The storage device 42 (database) stores, for each predetermined theme, evaluation points that express evaluation in quantitative terms, text data that express evaluation in free-form text using natural language, and data that expresses dates.

Figure 2 is a diagram showing an example of quantitative data (evaluation points) and qualitative data (text data). In Figure 2, dates, identifiers, evaluation points (quantitative data), and text data are associated with each other. This identifier is an identifier assigned to the quantitative data and qualitative data, that is, an identifier assigned to the combination of evaluation points and text data (comments such as customer requests and evaluations). Evaluation points are determined by the customer. In Figure 2, evaluation points are expressed as a five-level scale (five options), as an example: very positive "2", positive "1", average "0", negative "-1", and very negative "-2".

The analysis unit 43 shown in FIG. 1 acquires text data expressing evaluations in free text from the control unit 41. The analysis unit 43 performs an analysis process (text mining) on the text data expressing evaluations in free text. The analysis unit 43 classifies the text data expressing evaluations into text data expressing positive reviews, text data expressing average reviews (neither good nor bad), and text data expressing negative reviews. The analysis unit 43 classifies the text data expressing evaluations using a predetermined model. This predetermined model is not limited to a specific model, but is, for example, a model learned by a machine learning technique (supervised learning) using teacher data.

The analysis unit 43 executes a filtering process on the classified text data. For example, the analysis unit 43 extracts one or more elements (feature data) from the classified text data. These feature data are predetermined based on, for example, parameters (e.g., character strings of words, meanings of words) that are predetermined in the natural language processing function that is standardly installed in the analysis unit 43. For example, the analysis unit 43 extracts the character "red" from the text data based on a parameter representing the character "red". For example, the analysis unit 43 extracts a word that means a specific color (e.g., red, blue) from the text data based on a parameter representing the meaning "color". The analysis unit 43 may extract feature data that cannot be extracted by the standard natural language processing function alone from the text data using parameters (dictionary settings) specified by a user who operates the operation unit 20 of the user terminal 2. The feature data that cannot be extracted by the standard natural language processing function alone is, for example, a proper name such as a product name.

The analysis unit 43 performs clustering on the extracted elements (feature data) using a predetermined model. This predetermined model is not limited to a specific model, but is, for example, a model learned by an unsupervised learning method.

For example, the analysis unit 43 clusters elements that have similar characters or character strings. For example, the analysis unit 43 may cluster elements that have similar meanings. The clustered elements are represented, for example, in the form of a graph having clustering properties (for example, a scatter graph). The clustered elements may be represented, for example, in the form of a data table. The analysis unit 43 outputs the results of the clustering process for one or more elements to the control unit 41.

The analysis unit 44 obtains the analysis results (the results of the clustering process for one or more elements) by the analysis unit 43 from the control unit 41. That is, the analysis unit 44 obtains one or more clusters (elements classified into classes) from the control unit 41. The analysis unit 44 obtains evaluation points (weighting coefficients) that represent evaluations in quantitative terms from the control unit 41.

The analysis unit 44 combines one or more clusters with evaluation points (weighting coefficients) associated with the elements of each cluster. For example, the analysis unit 44 uses the evaluation points to perform weighting on the elements of each cluster. As a result, the analysis unit 44 selects each element of the requests and evaluations that have a greater impact on customer satisfaction from among the elements of one or more clusters. In other words, the analysis unit 44 selects one or more elements with high priority from among the elements of one or more clusters. The priority is, for example, the number of elements in a cluster. The priority may be, for example, the sum of the evaluation points of the elements in a cluster. The sum of the evaluation points may be the sum of the absolute values of the evaluation points. The analysis unit 44 outputs one or more elements with high priority to the creation unit 45.

The creation unit 45 acquires one or more high priority elements from the analysis unit 44. The creation unit 45 generates data representing measures for improving or maintaining customer satisfaction based on the one or more high priority elements. For example, the creation unit 45 creates a summary of text data (qualitative data) associated with the one or more high priority elements as data representing the measures. This summary is displayed on the display unit 23 (dashboard) as a sentence proposing the measures. This summary may be created using a template. The creation unit 45 may count the number of elements (number of cases). The creation unit 45 may also count the number of elements for each customer satisfaction level (evaluation point).

The creation unit 45 outputs data representing measures (proposal contents) for improving or maintaining customer satisfaction to the control unit 41. The data representing the measures is transmitted to the user terminal 2 by the communication unit 40. As a result, the data representing the measures is displayed on the display unit 23.

Next, the analysis unit 43, the analysis unit 44, and the creation unit 45 will be described in detail.
3 is a diagram showing an example of a classification result of qualitative data. The classification unit 430 acquires the identifier and the text data from the control unit 41 among the date, the identifier, the evaluation score (quantitative data), and the text data (qualitative data) in the data table shown in FIG.

The classification unit 430 classifies the text data representing evaluations into text data representing positive reviews, text data representing average reviews (neither good nor bad), and text data representing negative reviews. The classification unit 430 classifies the text data representing evaluations by text mining using a predetermined model.

In FIG. 3, the classification unit 430 classifies the text data "Because the design is good" associated with the identifier "ID-1" into "Well-received" based on, for example, the character string "good". The classification unit 430 classifies the text data "Because the material is good" associated with the identifier "ID-5" into "Well-received" based on, for example, the character string "good". The classification unit 430 classifies the text data "Because the material is durable" associated with the identifier "ID-7" into "Well-received" based on, for example, the character string "durable".

In FIG. 3, the classification unit 430 classifies the text data "The price is questionable" associated with the identifier "ID-3" as "normal" based on the character string "Very questionable," for example.

In FIG. 3, the classification unit 430 classifies text data "Because the size doesn't fit" associated with the identifier "ID-2" into "unpopular" based on, for example, the character string "doesn't fit". The classification unit 430 classifies text data "Because there is no favorite color (blue)" associated with the identifier "ID-4" into "unpopular" based on, for example, the character string "favorite color...none". The classification unit 430 classifies text data "Because I don't like red" associated with the identifier "ID-4" into "unpopular" based on, for example, the character string "don't like".

Figure 4 is a diagram showing an example of the extraction result of elements of the classification result. The extraction unit 431 executes a filtering process on the classified text data. For example, the extraction unit 431 extracts one or more elements (feature data) from the classified text data. These feature data are predetermined by the control unit 41 based on parameters specified by the data processing unit 21. The extraction unit 431 outputs the result of the filtering process to the clustering unit 432.

In FIG. 4, the extraction unit 431 extracts the element "design" from the text data based on the parameters representing the character string "design". The extraction unit 431 extracts the element "design" from the text data based on the parameters representing the character string "material". The extraction unit 431 extracts the element "price" from the text data based on the parameters representing the character string "price". The extraction unit 431 extracts the element "size" from the text data based on the parameters representing the character string "size". The extraction unit 431 extracts an element meaning a specific color (e.g., red, blue) from the text data based on the parameters (semantic tags) representing the meaning "color". The extraction unit 431 may extract the element "red" from the text data based on the parameters representing the character "red". As described above, these feature data are predetermined based on parameters (e.g., character strings of words, meanings of words) predetermined in the natural language processing function installed as standard in the extraction unit 431. In addition, the extraction unit 431 may use parameters (dictionary settings) specified by a user who operates the operation unit 20 of the user terminal 2 to extract feature data from the text data that cannot be extracted by the built-in natural language processing function alone.

Figure 5 is a diagram showing an example of the clustering result of the extraction result. The clustering unit 432 obtains the result of the filtering process (one or more extracted elements) from the extraction unit 431. The clustering unit 432 performs clustering on the one or more extracted elements (feature data).

The clustered elements are represented, for example, in the form of a graph with clustering (e.g., a scatter graph). The two orthogonal axes in the graph with clustering are determined, for example, as a result of dimensionality reduction on a plurality of extracted elements (feature data). In a graph with clustering, the more similar the characters or character strings are to each other, the shorter the distance between the elements. Also, in a graph with clustering, the closer the elements are to each other in meaning, the shorter the distance between them.

In FIG. 5, the clustering unit 432 clusters elements with similar characters or character strings. For example, the clustering unit 432 clusters the element "material" associated with the identifier "ID-5" and the element "material" associated with the identifier "ID-7". The clustering unit 432 may cluster elements with similar meanings. For example, the clustering unit 432 clusters the element "blue" associated with the identifier "ID-4" and the element "red" associated with the identifier "ID-6". The clustering unit 432 outputs the results of the clustering process for the elements to the control unit 41.

Figure 6 is a diagram showing an example of the display of the proposed content. The analysis unit 44 derives a priority for the elements of each cluster. For example, the number of elements in the cluster of the element "material" shown in Figure 5 is two. For example, the number of elements in the cluster of the element "blue" and the element "red" shown in Figure 5 is two. In such a case, for example, the priority of the element "material" and the priority of the element "color" are treated as having the same priority.

The analysis unit 44 may use the evaluation points to derive a priority for the elements of each cluster. For example, the sum of the evaluation points of the elements in the cluster of the element "material" shown in FIG. 5 is the sum of the evaluation point "1" and the evaluation point "1" shown in FIG. 2, which is "2". For example, the sum of the evaluation points of the elements in the clusters of the element "blue" and the element "red" shown in FIG. 5 is the sum of the evaluation point "1" and the evaluation point "-2" shown in FIG. 2, which is "-1". In such a case, the priority of each element may be treated as being higher the larger the sum (absolute value).

The creation unit 45 creates a summary of the text data (qualitative data) associated with one or more high-priority elements as data representing a measure. In FIG. 6, the creation unit 45 creates text data "Customers comment that the material is good. Let's continue using this material" as a summary of the text data "Because the material is good" which is a high-priority factor for favorable reviews, and the text data "Because the material is durable" which is a high-priority factor for favorable reviews.

In FIG. 6, the creation unit 45 creates text data "One customer comments that 'there is no favorite color'. Another customer comments that 'I don't like red'. Let's increase the color variations." as a summary of the text data with high priority as a cause of unpopularity, "Because there is no favorite color (blue)," and the text data with high priority as a cause of unpopularity, "Because I don't like red." These summary sentences are displayed, as an example, in the "Suggestion (Insight)" field on the display unit 23 as sentences proposing measures.

A graph (syntax information) with clustering properties is displayed, for example, in the "Cluster" column on the display unit 23. A pie chart showing the distribution of ratings is displayed, for example, in the "Customer Satisfaction" column on the display unit 23. A bar graph (word distribution) showing the number of elements of favorable reviews is displayed, for example, in the "Factors of favorable reviews" column on the display unit 23. A bar graph (word distribution) showing the number of elements of unfavorable reviews is displayed, for example, in the "Factors of unfavorable reviews" column on the display unit 23.

Next, an example of the operation of the proposed device 4 will be described.
7 is a flowchart showing an example of the operation of the proposal device 4. The control unit 41 acquires quantitative data and qualitative data obtained from multiple customers from the user terminal 2. The quantitative data is, for example, evaluation points. The qualitative data is, for example, text data (freely written comments) (step S101). The analysis unit 43 executes a predetermined analysis process on the qualitative data. The predetermined analysis process is, for example, a process of classifying the qualitative data, a process of extracting multiple elements from the classified qualitative data, and a clustering process on the extracted multiple elements (step S102).

The analysis unit 44 performs weighting processing on the results of a predetermined analysis processing using multiple quantitative data ranging from positive to negative (e.g., multiple evaluation points covering a range from positive to negative). For example, the analysis unit 44 performs weighting processing on the multiple extracted elements using the evaluation points (step S103). The creation unit 45 creates data representing measures related to the quantitative data based on the results of the weighting processing. Measures related to the quantitative data are, for example, measures related to improving or maintaining customer satisfaction (evaluation points). The creation unit 45 creates, for example, a summary of text data associated with one or more elements with high priority (weighting) as data representing the measures (step S104). The control unit 41 transmits the data representing the measures to the user terminal 2 using the communication unit 40 (step S105).

As described above, the control unit 41 acquires evaluation scores (quantitative data) and text data (qualitative data) obtained from multiple customers from the user terminal 2. The analysis unit 43 executes a predetermined analysis process on the qualitative data. The analysis unit 44 uses the quantitative data to execute a weighting process on the results of the predetermined analysis process. The creation unit 45 creates data representing measures related to the quantitative data based on the results of the weighting process.

This makes it possible to improve the accuracy of measures to improve or maintain customer satisfaction. It makes it possible to improve the accuracy of measures to improve or maintain key performance indicators.

It is difficult for a person to comprehensively analyze a large amount of quantitative data and qualitative data, for example, in terms of time and low cost. In contrast, in the first embodiment, the proposal device 4 comprehensively analyzes (cross-analyzes) and analyzes a large amount of quantitative data and qualitative data, making it possible to propose highly accurate measures based on the true causes of improvement or decrease in customer satisfaction. Such a large amount of quantitative data (evaluation points) comprehensively includes multiple evaluation points ranging from good to bad. In addition, it is possible to output measures necessary to improve or enhance key performance indicators in a short time and at low cost. Furthermore, the person in charge of the company's services (analyst) can quickly improve or enhance the company's services and products based on the measures necessary to improve or enhance key performance indicators.

In the first embodiment, it is possible to visualize quantitative data and qualitative data in a format that is easy for the person in charge of the company's services to understand. For example, it is possible to visualize the word distribution of the qualitative data, the syntactic information of the qualitative data, the results of positive/negative analysis, and the clustering results of the qualitative data. Furthermore, by using the quantitative data (evaluation points) as, for example, a weighting coefficient for the analysis results of these qualitative data, the person in charge can obtain visualized analysis results that directly lead to the improvement or refinement of the company's services.

In FIG. 4, the text data "Because I don't like red" is classified as "unpopular" based on the results of text mining. However, in FIG. 2, the evaluation score associated with the text data "Because I don't like red" is a favorable score of "1." As in this example, even if a survey result would be classified as "unpopular" using text mining alone, there are cases where the survey result obtained from the customer is not actually unpopular (evaluation score of "1"). Even in such cases, the analysis unit 44 performs analysis processing using not only the text data (qualitative data) but also the evaluation score (quantitative data), making it possible to propose measures with high accuracy.

Second Embodiment
The second embodiment differs from the first embodiment in that the proposal device 4 acquires time-series evaluation points (quantitative data) and text data (qualitative data) representing the time-series evaluation. The second embodiment will be described focusing on the differences from the first embodiment.

The user terminal 2 collects the evaluation scores in a time series and text data representing the evaluations in a time series. The control unit 41 manages this data in a time series according to the date of collection.

Figure 8 is a diagram showing an example of quantitative data (evaluation points) and qualitative data (text data). In Figure 8, multiple different dates "12/15", "3/25", and "6/25" are associated with an identifier, an evaluation point (quantitative data), and text data.

The analysis unit 43 shown in FIG. 1 acquires text data representing the time-series evaluations from the control unit 41. The analysis unit 43 performs an analysis process (text mining) on the text data representing the time-series evaluations.

The analysis unit 43 performs a filtering process on the classified text data for each date. For example, the analysis unit 43 extracts one or more elements (feature data) from the classified text data for each date.

The analysis unit 43 performs clustering on the extracted elements (feature data) for each date. The analysis unit 43 outputs the results of the clustering process on the elements for each date to the control unit 41.

The analysis unit 44 obtains the analysis results (results of the clustering process on the elements) by the analysis unit 43 from the control unit 41 for each date. That is, the analysis unit 44 obtains one or more clusters (elements classified into classes) from the control unit 41 for each date. The analysis unit 44 obtains an evaluation score (weighting coefficient) that expresses the evaluation in quantitative terms from the control unit 41 for each date.

The analysis unit 44 combines one or more clusters with evaluation points (weighting coefficients) associated with the elements of each cluster for each date. For example, the analysis unit 44 uses the evaluation points to weight the elements of each cluster for each date. In this way, the analysis unit 44 selects each of the request and evaluation elements that have a greater impact on customer satisfaction from among the elements of one or more clusters for each date. In other words, the analysis unit 44 selects one or more elements with high priority from among the elements of one or more clusters for each date.

Figure 9 is a diagram showing an example of time-series key performance indicators. The horizontal axis indicates the dates on which the quantitative data and qualitative data were collected. The vertical axis indicates the key performance indicators. The calculation method of the key performance indicators is predetermined. For example, the key performance indicators are calculated for each date based on one or more customer satisfaction levels. The analysis unit 44 calculates the key performance indicators based on the predetermined calculation method. In Figure 9, the key performance indicator dated December 25th is "2" as an example. The key performance indicator dated March 25th is "3" as an example. The key performance indicator dated June 25th is "4" as an example.

The creation unit 45 acquires the time-series key performance indicators from the analysis unit 44. The creation unit 45 generates a graph representing the time-series key performance indicators based on the acquired time-series key performance indicators. The graph representing the time-series key performance indicators is displayed on the display unit 23 (dashboard).

Figure 10 is a diagram showing an example of the display of the proposed content. The creation unit 45 creates a summary of the text data (qualitative data) associated with one or more high-priority elements as data representing the measure. In Figure 10, the creation unit 45 creates the text data "Customers comment that 'because you can choose your favorite color'. Let's continue offering a variety of colors." as a summary of the text data with high priority as a factor of favorable reviews, "Because they have your favorite colors," and the text data with high priority as a factor of favorable reviews, "Because you can choose your favorite color."

In FIG. 10, the creation unit 45 creates text data "Customers comment that it's because the size doesn't fit. Let's increase the variety of sizes" as a summary of the text data "Because the size doesn't fit" which has a high priority as a reason for unpopularity. These summary sentences are displayed, as an example, in the "Proposal (Insight)" field on the display unit 23 as sentences proposing measures. In addition, a graph showing a time series of key performance indicators is displayed, as an example, in the "Key Performance Indicators" field on the display unit 23.

The person in charge may operate the operation unit 20 to specify or select a date (e.g., "12/25", "3/25", or "6/25") for the proposal contents, etc., to be displayed on the display unit 23. As a result, the "Proposal (Insight)" column displays the proposal contents associated with the specified or selected date. In addition, the "Customer Satisfaction (Numerical Evaluation)" column, the "Factors of Positive Evaluation" column, the "Factors of Negative Evaluation" column, the "Positive Comments (Free Text)" column, and the "Negative Comments (Free Text)" column each display the analysis results associated with the specified or selected date.

As described above, the control unit 41 acquires time-series quantitative data and time-series qualitative data. The quantitative data is, for example, evaluation points. The qualitative data is, for example, text data (freely written comments). The analysis unit 43 performs a predetermined analysis process on the time-series qualitative data. The analysis unit 44 performs a time-series weighting process using multiple quantitative data ranging from positive to negative (for example, multiple evaluation points covering a range from positive to negative). The creation unit 45 creates data representing measures related to the time-series quantitative data based on the results of the time-series weighting process.

This makes it possible to improve the accuracy of measures to improve or maintain customer satisfaction based on the true causes of trends in customer satisfaction. It makes it possible to improve the accuracy of measures to improve or maintain key performance indicators based on the true causes of trends in customer satisfaction. It makes it possible to manage over time how key performance indicators have improved through improvement activities.

In the second embodiment, it is possible to visualize time-series quantitative data and qualitative data in a format that is easy for the person in charge of the company's services to understand. For example, it is possible to visualize the word distribution of the qualitative data, the syntactic information of the qualitative data, the results of positive/negative analysis, the clustering results of the qualitative data, and the trends of the qualitative data.

(Hardware configuration example of the proposed device 4 in each embodiment)
11 is a diagram showing an example of a hardware configuration of the proposed device 4 in each embodiment. Some or all of the functional units of the proposed device 4 are realized as software by a processor 100 such as a CPU (Central Processing Unit) executing a program stored in a storage device 42 having a non-volatile storage medium (non-transient storage medium). The program may be recorded in a computer-readable storage medium. The computer-readable storage medium is, for example, a portable medium such as a flexible disk, a magneto-optical disk, a ROM (Read Only Memory), a CD-ROM (Compact Disc Read Only Memory), or a non-transient storage medium such as a storage device such as a hard disk built into a computer system.

Some or all of the functional units of the proposed device 4 may be realized using hardware including electronic circuits (electronic circuits or circuitry) using, for example, LSIs (Large Scale Integrated circuits), ASICs (Application Specific Integrated Circuits), PLDs (Programmable Logic Devices), FPGAs (Field Programmable Gate Arrays), etc.

The present invention is applicable to systems that provide text mining, text analysis, or voice text log analysis services.

The above describes an embodiment of the present invention in detail with reference to the drawings, but the specific configuration is not limited to this embodiment, and includes designs that do not deviate from the gist of the present invention.

1...Proposed system, 2...User terminal, 3...Communication line, 4...Proposed device, 20...Operation unit, 21...Data processing unit, 22...Communication unit, 23...Display unit, 40...Communication unit, 41...Control unit, 42...Storage device, 43...Analysis unit, 44...Analysis unit, 45...Creation unit, 100...Processor, 101...Memory

Claims (5)

A control unit that acquires evaluation points and text data obtained from a plurality of customers;
an analysis unit that executes a predetermined analysis process on the text data;
an analysis unit that performs a weighting process on the result of the predetermined analysis process using a plurality of the evaluation points ranging from positive to negative;
a creation unit that creates data representing measures related to the evaluation points based on a result of the weighting process ,
the analysis unit executes, as the predetermined analysis process, a process of classifying the text data, a process of extracting a plurality of elements from the classified text data, and a clustering process for the extracted plurality of elements.
Proposed device. The analysis unit performs a weighting process on the plurality of elements,
the creation unit prioritizes the element having a large weighting, and creates data representing a measure regarding the evaluation point associated with the prioritized element.
The proposal device according to claim 1 . The control unit acquires the evaluation score in time series and the text data in time series,
The analysis unit executes the predetermined analysis process on the time-series text data,
The analysis unit performs the weighting process on the time series using the evaluation points on the time series;
The creation unit creates data representing measures related to the evaluation points in a time series based on a result of the weighting process in a time series.
The proposal device according to claim 1 or 2 . A proposal method executed by a proposal device,
A control step of acquiring evaluation points and text data obtained from a plurality of customers;
an analysis step of performing a predetermined analysis process on the text data;
an analysis step of performing a weighting process on the result of the predetermined analysis process using a plurality of the evaluation points ranging from positive to negative;
and creating data representing measures related to the evaluation points based on a result of the weighting process ,
The analysis step includes executing, as the predetermined analysis process, a process of classifying the text data, a process of extracting a plurality of elements from the classified text data, and a clustering process for the extracted plurality of elements.
Proposal method. A program for causing a computer to function as the proposal device according to any one of claims 1 to 3 .

Images