JP7398320B2 - Consulting support system, consulting support method - Google Patents

Description

The present invention relates to a consulting support system and a consulting support method.

Conventionally, there have been various techniques for consulting users using topic models. For example, in Patent Document 1, a topic model is used to analyze user behavior on a certain website (domain) and information about that site to recommend products and services, and the characteristics of the domain analyzed using the topic model are used to recommend products and services. It can also be used for domains.

Japanese Patent Application Publication No. 2016-76097

In Patent Document 1, behavior prediction accuracy is improved by extracting different user profiles for each domain group, and features in a domain analyzed using a topic model can be utilized in other domains. However, in order to overcome the differences in definitions of words in different domains, it is necessary to match the definitions of words in each domain. With this method, a topic model must be created each time another domain is visited. In other words, word dictionaries are ad-hoc, and by accumulating and utilizing their analysis and knowledge, that is, feeding them back, how can we resolve differences in word definitions in the domain when performing analysis using topic models? The idea was not disclosed. Furthermore, when the former domain owner introduced a recommendation system to the new owner of another domain, in the upstream process, he used a dictionary that resolved differences in definitions between words obtained in the domain where recommendations had been introduced in the past. Unable to provide consulting.

One aspect of the present invention is to provide a consulting support system and a consulting support method that are capable of resolving differences in definitions of words in a domain and performing analysis using a topic model.

The consulting support system according to the present invention uses a user topic model obtained from user data including user public data and user private data owned by a user company, and external public information published outside other than the user company. a topic model generation unit that generates an external topic model obtained from data and a mixed topic model that serves as a reference for determining the degree of association between the external topic model and the user topic model; A feedback topic is identified from the user topic model, which is a topic for which the distance between the topic owned by the user topic model and the topic owned by the user topic model is equal to or less than a threshold value, and the topic owned by the mixed topic model and the topic owned by the external topic model are identified. a topic model analysis unit that identifies an external topic from the external topic model, which is a topic for which a distance of It is configured as a consulting support system characterized by having a department and a.

According to one aspect of the present invention, it is possible to resolve differences in definitions of words in a domain and perform analysis using a topic model.

1 is a diagram illustrating a configuration example of a consulting support system according to the present embodiment. FIG. 3 is a diagram illustrating an example of government agency open data included in externally disclosed data. FIG. 2 is a diagram showing an example of construction machinery export amount data. FIG. 3 is a diagram showing an example of coal production amount transition data. It is a figure showing an example of sales data of a user company. It is a diagram showing an example of a feedback repository (initial state). FIG. 3 is a diagram showing an example of a feedback repository (after update). It is a flowchart which shows the processing procedure of analysis processing. 9 is a flowchart showing detailed steps of the processing of S812 and S820 shown in FIG. 8. 9 is a flowchart showing detailed steps of the process of S824 shown in FIG. 8. 9 is a flowchart showing detailed steps of the processing from S815 to S819 shown in FIG. 8. FIG. 6 is a diagram showing an example of a screen on which the analysis setting section displays the analysis results on the display section. FIG. 7 is a conceptual diagram for explaining the processing of S815 to S819. FIG. 2 is a conceptual diagram for explaining topic models and data collected by a related data collection unit, and their relationships; FIG.

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of a consulting support system and a consulting support method according to the present invention will be described in detail below with reference to the accompanying drawings.

FIG. 1 is a diagram showing a configuration example of a consulting support system in this embodiment. As shown in FIG. 1, the consulting support system 1000 uses a user server 200, which is a computer managed by a user company undergoing consulting, and external public data 300, which is made public outside the user company. It has a consulting server 100 for providing consulting to the above user companies. The consulting server 100 is connected to a user server 200 via a network N1, and is connected to a cloud or a server in which external public data 300 is stored via a network N2. The following explanation assumes that each process is executed by the consulting server 100, but one or more servers may execute each process, or a cloud having the same functions as the consulting server 100 can execute each process. may be executed.

Each process in the consulting server 100 is realized by executing a program. The consulting server 100 uses general memory such as a CPU (Central Processing Unit), ROM (Read Only Memory), RAM (Random Access Memory), HDD (Hard Disk Drive), and NIC (Network Interface Card) that are included in a communicable computer. The CPU reads a program from the ROM and executes it, thereby realizing the functions of the functional unit that executes each process. The above program may be provided to the consulting server 100 by being read from a storage medium such as a USB (Universal Serial Bus) memory or by being downloaded from another computer via a network.

The consulting server 100 is a server for receiving a request (for example, a problem solving request) from a user server 200 owned by a user company and presenting a solution to the problem, and is configured of a general computer as hardware. be done. The networks N1 and N2 are composed of general public line networks such as the Internet, regardless of whether they are wired or wireless.

The consulting server 100 includes a user data acquisition unit 101 for acquiring various user data owned by user companies used in this system from the user server 200, an external data acquisition unit 102 for acquiring external public data 300, and Using the data and the feedback repository 109 described later, an analysis process is performed on the user company based on a predetermined analysis method (for example, an analysis method for formulating a business strategy such as SWOT analysis or five force analysis), and the results are an analysis setting section 103 that displays on the display section 104; an alert processing section 105 that outputs alert information when the analysis result by the analysis setting section 103 exceeds a predetermined threshold; The related data collection unit 106 collects the topic model and its source data) from the topic model analysis unit 111, and the visualization data that visualizes the data collected by the related data collection unit 106 and the topic name of the topic model are determined. A related data display control unit 107 for generating and outputting keyword display data (hereinafter referred to as provision data) for displaying keywords serving as main words and phrases, and provision data generated by the related data display control unit 107. and a data transmitting unit 108 that transmits the data to the user server 200.

As will be specifically described later, user data includes user public data, which is data disclosed to the public by the user company, and user private data, which is data not disclosed to the public by the user company. Below, we will explain using IR (Investor Relations) data 2011 regarding business conditions, financial conditions, and performance trends (for example, trends in financial results data) that user companies transmit to investors as an example of user public data. However, in addition to this, the user public data may include various data disclosed to the public by each user company, including the user company's annual report 2012. IR Data 2011 includes management indicators showing the user company's business status, financial situation, and business performance trends, as well as the name of the industry and industry to which the user company belongs, financial results data of the user company, and the number of stores and branches operated by the user company. , management goals of the user company, and various other data that the user company discloses on a regular or irregular basis. For example, IR Data 2011 includes the user company's sales by country (such as the sales amount of products in each country, such as North America: 130 billion yen, China: 120 billion yen, etc.), and Annual Report 2012 includes, for example, Management goals of the user company are included, such as sales business target amount, rental business target amount, etc. for each construction machine.

Further, in the following description, sales data 2023 of the user company, which is the basis of the IR data 2011, will be used as an example of user-disclosed data. Accounting data 2022 related to the accounting of the user company, customer data 2024 showing the product sales customers of the user company, inventory data 2025 related to the inventory status of the products manufactured by the user company, and product data 2026 showing the manufacturing status of the products manufactured by the user company. First, the user private data may include various data that is the source of the user public data and that is not generally disclosed by each user company.

The consulting server 100 further generates a user topic model 1111 from the various user data acquired by the user data acquisition unit 101, generates an external topic model 1112 from the external public data acquired by the external data acquisition unit 102, and generates an external topic model 1112 from the external public data acquired by the external data acquisition unit 102. A topic model generation unit 112 generates a mixed topic model 1113 from the data and the external public data, and a topic model generation unit 112 determines the relevance of items in each topic model generated, and determines the items that meet predetermined conditions. , and a feedback learning unit 110 that adds the topic names extracted by the topic model analysis unit 111 to a feedback repository 109, which will be described later. Each server included in this system and specific processing executed by each of these units will be described later. Note that the arrows mainly indicate the flow of data.

In the following, explanation will be given using the government open data 302 that is disclosed to the public by each government agency as an example of externally disclosed data. Weather data from all over the country published by 303, credit investigation data 304 related to corporate credit investigations, research reports 305 such as government statistics, news article data 306 containing various news related to the government and each company, etc. Externally disclosed data may include various data disclosed to the public by governments and companies other than companies. Next, the data used in this system will be explained. First, the external public data 300 will be explained.

FIG. 2 is a diagram showing an example of government agency open data 302 included in external public data 300. The ministries and agencies open data 302 is data that each ministry and agency makes available to the public. FIG. 2 shows data 3021 regarding Japan's export amount of construction machinery (hereinafter referred to as construction machinery export amount data) as an example of the ministry open data 302. Figure 2 shows the amount of construction machinery exported from Japan to Asia and other regions in each of the years 2007, 2012, and 2017.

Further, as shown in FIG. 3, the construction machinery export value data 3021 includes data 30211 indicating changes in the shipment amount for each construction machine (hereinafter referred to as shipment trend data for each machine). In Figure 3, it can be seen that for each year of 2007, 2012, and 2017, the domestic shipment amount, export amount, and subtotal of these are shown for each construction machine such as tractors and hydraulic excavators.

In addition, as shown in FIG. 4, the ministry open data 302 includes data 3022 indicating changes in world coal production (hereinafter referred to as coal production change data). In Figure 4, it can be seen that for each year from 2005 to 2015, the amount of coal produced in countries around the world, including China and the United States, and the total for each year are shown. Next, user data will be explained.

FIG. 5 shows sales data 2023 of a user company as an example of user public data included in user data. In Figure 5, a number (#) indicating sales information, the name of the sold product recorded in the sales information indicated by the number, a product category indicating the product category, and the number of customers who purchased the product in the product category are shown. It can be seen that the customer name and the type of business or industry in which the customer is engaged are stored in association with each other. For example, in Figure 5, the customer with number 1 (#1) is "High-rise Building Construction Co., Ltd.", a domestic company whose industry is in the construction business, and whose product category is "crane", which is the product "Crane Type A". ” was purchased and the sales were recorded. Sales data 2023 indicates the basis and breakdown of the user company's sales data stored as IR data 2011. As will be described later, when a user topic model is generated, the product categories "crane" and "bulldozer" are subjected to topic model processing as topic names "building construction machinery" and "excavation machinery," respectively. Next, the feedback repository 109 held by the consulting server 100 will be explained.

6 and 7 are diagrams showing examples of the feedback repository 109. The feedback repository 109 absorbs and resolves differences in the definitions of words used in each domain published as external public data, and is used to associate topics in the external topic model 1112 with topics in the user topic model 1111. It is a database. FIG. 6 is an example of the feedback repository 109a before being updated in the process described below, and FIG. 7 is an example of the feedback repository 109b after being updated in the process described later.

As shown in FIGS. 6 and 7, the feedback repository 109 contains the serial number (#) of the record, the external topic name that is the topic name defined in the external topic model 1112, and the topic name defined in this repository. A feedback topic name and a data domain indicating the user data that is the source of the feedback topic name and the type of business or industry to which the user company belongs are stored in association with each other. Since FIG. 6 shows the feedback repository 109a in its initial state, the same content as the external topic model 1112 is stored in each of these items. In the feedback repository 109b shown in FIG. 7, the topic name of the user topic model 1111 is newly added as a feedback topic name in association with the external topic name by executing the process described below. In FIG. 7, for example, it can be seen that the topic name "Building Construction Machinery" of the user topic model 1111 is stored as a serial number "n.1" in association with the external topic name "Construction Machinery". Further, as the added feedback topic name, the data domain name of the topic belonging to the topic model related to the construction industry in the user topic model 1111 is stored. The specific process will be described later, but by storing and storing such relationships in the feedback repository 109, the definition of the words used in the user data and the words with the same meaning. In addition to absorbing and resolving differences in word definitions in externally published data, where different words and phrases are used depending on the type of business or industry, the items included in the topics of the user topic model are changed from what users have previously noticed. It is possible to realize that items included in the topics of the external topic model, which were not previously available, are related.

FIG. 8 is a flowchart showing the processing procedure of analysis processing performed in this system when consulting a user company. As shown in FIG. 8, the external data acquisition unit 102 first acquires external public data 300 as shown in FIG. 1, copies it to an internal memory (not shown), and stores the external public data 300 ( S801). This allows information published by governments, ministries, and research institutions to be accumulated as external data. For example, the external public data 300 is stored as government CIO data 301a, ministry open data 302a, weather data 303a, credit investigation data 304a, investigation report 305a, and news article data 306a.

Next, the user data acquisition unit 101 acquires user data as shown in FIG. 1 from the user server 200, copies user public data included in the user data to an internal memory (not shown), and stores it (S802). ). At this time, the user data acquisition unit 101 obtains from the IR data 2011 and the annual report 2012, in addition to the management indicators of the user company, the name of the industry and industry to which the user company belongs, the sales data and financial results data of the user company, and the user company's information. Read various data such as the number of stores and branches. For example, user public data is stored as IR data 2011a and annual report 2012a.

Further, the user data acquisition unit 101 determines whether the user data includes user private data that is internal data of the user company (S803), and determines whether the user data includes user private data. If it is determined (S803; Yes), the user private data is copied to an internal memory (not shown) and stored (S804). For example, the user data acquisition unit 101 stores sales data of the user company, which is the basis of the IR data 2011 and the annual report 2012, as the user private data. On the other hand, if the user data acquisition unit 101 determines that the user data does not include user private data (S803; No), the process proceeds to S805 without doing anything.

Upon completion of S803 or S804, the topic model generation unit 112 generates a user topic model 1111 from the user data (if user private data is included, the user data including the user private data) (S805), and generates a topic model It is registered in the analysis unit 111 (S806). As a method for generating a user topic model, conventionally known methods such as LDA (Latent Dirichlet Allocation), PLSI (Probabilistic Latent Semantic Indexing), and LSI (Latent Semantic Index) can be used.

Furthermore, when executing S804, the topic model generation unit 112 generates a mixed topic model 1113 from the user data and externally published data (S807), and registers it in the topic model analysis unit 111 (S808). As for the method of generating the mixed topic model, a conventionally known method can be used as in the case of generating the user topic model. The mixed topic model is a user topic model 1111 that serves as a standard for determining the semantic similarity between the topic name of the user topic model 1111 and the topic name of the external topic model 1112 based on the distance between items of the topic model. This is a topic model that is commonly used between the user topic model 1111 and the external topic model 1112, and is a topic model that is used within this system to determine whether there is a relationship between the user topic model 1111 and the external topic model 1112.

In S801, when the external data acquisition unit 102 acquires external public data 300 and stores it in an internal memory (not shown), the topic model generation unit 112 generates an external topic model 1112 from the external public data (S809), and It is registered in the model analysis unit 111 (S810). As for the method of generating the external topic model, conventionally known methods can be used, as in the case of generating the user topic model and the mixed topic model.

When each topic model is registered, the analysis setting unit 103 defines items that serve as analysis indicators for user data as internal factors (S811), and defines items that serve as analysis indicators for externally published data as external factors (S812). ). These definitions are, for example, defined by the analysis setting unit 103 when an analyst specifies items to be held as user data or externally disclosed data from an input device (for example, a keyboard) provided in the consulting server 100.

Once these items are defined, the analysis setting unit 103 determines whether the feedback repository 109 has been created (S813), and if it is determined that the feedback repository 109 has not been created (S813; No), In order to provide consulting to the user company, an analysis method for formulating a business strategy, such as SWOT analysis or five force analysis, is selected, and a threshold value to be used in the analysis method is defined (S814). The analysis method and threshold settings are defined, for example, by the analysis setting unit 103 having an analyst designate items to be held as user data or external public data from the input device included in the consulting server 100. Thereafter, the process proceeds to S821, and the analysis setting unit 103 performs analysis using the predetermined analysis method set in S814 and the threshold value used in the analysis method, with the internal factors defined in S811 and the external factors defined in S812 as axes. I do.

On the other hand, if the analysis setting unit 103 determines that the feedback repository 109 has been generated (S813; Yes), it reads the feedback repository 109 and redefines the items that serve as analysis indicators for the externally published data defined in S812. (S820), proceed to S814. In this case, as in the case described above, the analysis setting unit 103 selects an analysis method for consulting the user company, defines the threshold value used in the analysis method (S814), and then uses the internal factors defined in S811. Analysis is performed using the predetermined analysis method set in S814 and the threshold used in the analysis method, with the external factors redefined in S820 as the axis. Processing related to generating the feedback repository 109 will be described below.

In S806, S808, and S810, when the topic model generation unit 112 generates the user topic model 1111, external topic model 1112, and mixed topic model 1113, the topic model analysis unit 111 calculates the distance between items between each topic model. A topic name having an item whose value is less than or equal to a predetermined threshold is identified (S815). The distance between the items can be calculated using, for example, various conventionally known cluster analyses.

The topic model analysis unit 111 determines whether the distance between the topic name of the mixed topic model and the keyword that is the origin of the topic name is less than or equal to the threshold, among the topics that include items for which the distance between the items is less than or equal to a predetermined threshold. A topic name (for example, a topic name that includes a phrase indicating the characteristics of the topic) is specified, and a topic name of the user topic model 1111 corresponding to the specified topic name is extracted (S816). The corresponding topic name is, for example, a topic name synonymous with the topic name or keyword of the identified mixed topic model 1113 in the user topic model 1111.

In addition, the topic model analysis unit 111 determines that, among topics that include items for which the distance between the items is equal to or less than a predetermined threshold, the distance between the topic name of the mixed topic model and the keyword that is the source of the topic name is A topic name that is equal to or less than a threshold value (for example, a topic name that includes a word or phrase indicating the characteristics of the topic) is identified, and the topic name of the external topic model 1112 corresponding to the identified topic name and the topic name are extracts topic names of different external topic models 1112 (S817). The corresponding topic name is, for example, a topic name synonymous with the topic name or keyword of the identified mixed topic model 1113 in the external topic model 1112. Further, the different topic name is, for example, a topic name in the external topic model 1112 that is semantically similar to a topic name of the identified mixed topic model 1113 by a certain level or more, other than the synonymous topic name. For determining whether or not the two words are semantically close to each other by more than a certain level, various conventionally known semantic analysis methods such as PLSA (Probabilistic Latent Semantic Analysis) may be used. Here, the processing of S815 to S819 will be specifically explained.

FIG. 13 is a conceptual diagram for explaining the processing of S815 to S819. In S815, the topic model analysis unit 111 calculates the distance between the items included in the topics included in the mixed topic model 1113 and the items included in the topics included in the user topic model 1111, and the distance between the items included in the topics included in the mixed topic model 1113, based on the mixed topic model 1113 (not shown). The distances between the items included in the topic model 1113 and the items included in the external topic model 1112 are calculated, and the topic name of the topic containing the item for which the calculated distance is less than or equal to a predetermined threshold is calculated. Specify from among the mixed topic models 1113. As a result, topics of the user topic model 1111 and topics of the external topic model 1112 that include items semantically similar to the topics of the mixed topic model 1113 can be specified based on the content of the items included in the mixed topic model 1113.

Then, in S816, the topic model analysis unit 111 selects the topic name of the mixed topic model 1113 and the keyword that is the origin of the topic name, among the topics that include items for which the distance between the items is equal to or less than a predetermined threshold. A topic name whose distance from Extract. In FIG. 13, as an example, the topic names of the country sales topic 1111a, the excavation machinery topic 1111b, the building construction machinery topic 1111c, and the mining topic 1111d are topics of the user topic model 1111 that correspond to the topic names for which the distance is equal to or less than the threshold value. Extracted as a name.

Furthermore, in S817, the topic model analysis unit 111 selects the topic name of the mixed topic model and the keyword that is the source of the topic name, among the topics that include items for which the distance between the items is equal to or less than a predetermined threshold. The topic name of the external topic model 1112 that corresponds to the identified topic name and the corresponding A topic name of external topic model 1112 that is different from the topic name is extracted. In FIG. 13, as an example, the topic names of energy (mined amount) topic 1112b, export equipment topic 1112c, and export country topic 1112d are extracted as the topic names of the external topic model 1112 corresponding to the topic names for which the distance is less than or equal to the threshold value. has been done. Furthermore, in this step, the topic model analysis unit 111 extracts a topic name of the external topic model 1112 that is different from the extracted topic name. In FIG. 13, as an example, it is determined that the topic name of the energy (power generation amount) topic 1112a is semantically similar to the topic name of the energy (mined amount) topic 1112b by more than a certain level, and the different external topic model 1112 is extracted as the topic name.

The feedback learning unit 110 converts the topic name of the user topic model 1111 extracted in S816, the topic name of the external topic model 1112 extracted in S817, and the topic name of the different external topic model 1112 into mutually related topics. and is added to the feedback repository 109 in association with the above (S818). For example, as shown in FIG. 7, the feedback learning unit 110 determines that the topic name of the mining topic 1111d extracted in S816 and the energy (mining amount) topic 1112b extracted in S817 are the same as in the mixed topic model 1113 identified above. Since the topic name is synonymous with the topic name or keyword, it is determined that the topics are related to each other. Then, the feedback learning unit 110 stores the topic name "Mining" of the mining topic 1111d in the feedback topic name, and adds the topic name "Energy" of the energy (mined amount) topic 1112b to the external topic name corresponding to the feedback topic name. (Mined amount)” is memorized. In Figure 7, the external topic name and the feedback topic name “Energy” (sequential number “p”) are set as the next record (sequential number “p.1”) after the record with the feedback topic name “Energy”. )", and a record in which "mining" is set as the feedback topic name is stored as an alias record of the external topic name of the above record (serial number "p"). Similarly, in FIG. 7, as the next record (serial number "n.1") of the record (serial number "n") with the initial external topic name and feedback topic name "Construction Machinery", the external topic name is " A record in which "Construction Machinery" and "Building Construction Machinery" are set as the feedback topic name is stored as an alias record of the external topic name of the above record (serial number "n").

Furthermore, the feedback learning unit 110 determines that the topic name "Energy (amount of power generation)" is semantically similar to "Energy (amount of mining)" set as an external topic name, and ) is extracted as the topic name of the different external topic model 1112, and is stored as an alias record (serial number "p.2") of the external topic name of the record (serial number "p"). Regarding the feedback topic name "Fuel" of the record, as described above, the topic name of the topic of the user topic model 1111 that is related to the energy (power generation amount) topic 1112a (topic name "Fuel" not shown in FIG. 13) is stored. ing.

After that, the feedback learning unit 110 sets the name of the business type or industry to which the user company and user data included in the IR data 2011 belong to the data domain of the added record (S819). The data domain may be set as the above name by, for example, the feedback learning unit 110 analyzing the type of business or industry from the contents of the IR data 2011 and user data (for example, title name, item name).

In this manner, when the processes of S815 to S819 are completed, the feedback repository 109a in the initial state shown in FIG. 6 is updated to the feedback repository 109b after learning shown in FIG. 7. When S819 ends, the process advances to S820, and uses the topic of the external data topic name and the topic model of the feedback topic name included in the updated feedback repository 109 to determine items that serve as analysis indicators for the external public data defined in S812. is redefined.

Then, the analysis setting unit 103 performs analysis using the predetermined analysis method set in S814 and the threshold value used in the analysis method (S821), and displays the analysis result on the display unit 104 such as a display (S822). .

FIG. 12 is a diagram showing an example of a screen on which the analysis setting section 103 displays the above analysis results on the display section 104. As shown in FIG. 12, the analysis setting unit 103 displays, on the display unit 104, the company goals 1202 of the user company and the results of an analysis method such as SWOT based on internal and external factors as analysis results 1201. An analysis table 1203 is displayed. The company goal 1202 may be read, for example, from the user public data of the user data acquired by the user data acquisition unit 101.

Further, the analysis setting unit 103 determines whether or not there are any items that do not satisfy the threshold set in S814 as a result of the above analysis. If the analysis setting unit 103 determines that there is an item that does not satisfy the threshold set in S814, the analysis setting unit 103 determines that the item has room for improvement for the user company, and the alert processing unit 105 notifies the user of an alert. At the same time, the corresponding item among the analysis results displayed in S822 is made selectable on the display unit 104 (S823). The above-mentioned alert is, for example, a message outputted by the alert processing unit 105 on the screen on which the above-mentioned analysis result is displayed in order to prompt a warning for an item that does not satisfy the above-mentioned threshold value.

In Figure 12, the analysis results show that in overseas sales, which was analyzed as a strength of the user company, the amount of construction equipment exports to China, which was analyzed as a threat to the user company, was less than a threshold (for example, ±0 compared to the previous year) -1 .4%. Therefore, in order to make the selection possible, the alert processing unit 105 uses, for example, the topic model and original data used to display the item as an analysis result from the item (the amount of construction equipment exported to China). The item 1204 is set to be accessible to the source data, such as by creating a link to the item 1204, and the screen after setting is displayed on the display unit 104.

Here, when the person in charge of the user company receiving consulting from the analyst selects the item 1204 displayed on the display section 104 and the alert processing section 105 accepts the selection, the alert processing section 105 selects the item 1204 displayed on the display section 104. 106, and the related data collection unit 106 collects related data that is the basis for the selected item 1204 (S824). For example, the related data collection unit 106 collects from the topic model analysis unit 111 the topic model used to output the item 1204 for which the alert was output as an analysis result and the data that is the source of the topic model. The collected topic model and the above source data are, for example, the external topic model 1112 used to output the above item 1204 (for example, export equipment topic 1112c, energy (mined amount) topic 1112b, energy (power generation amount) topic 1112a) ), their source data (for example, shipment trend data by machine 30211, coal production trend data 3022, power generation data by energy by country 3033), and user topic models 1111 (for example, mining topic 1111d, building construction machine topic 1111c, excavating machine topic 1111b) and the data that became the source of these (for example, sales data 2023, customer data 2024, product data 2026). In S820, if the analysis setting unit 103 redefines the items that serve as analysis indicators for the externally published data defined in S812, the collected topic models include topic models that include the redefined items and the relevant topic models. Contains the data that is the basis of the topic model.

The related data display control unit 107 displays the related topic models and the source data collected by the related data collection unit 106 and the relationships between the topic models and the source data on the display unit by, for example, switching screens. 104, and also outputs and displays the keyword that is the source of the topic name that is the basis for the relevance (S825). The related data display control unit 107 reads, as the keyword, a phrase that is the source of the topic name of the mixed topic model, or reads a phrase that is synonymous with the word and phrase that is the topic name of the external topic model 1112 or user topic model 1111. All you have to do is read the original phrase.

FIG. 14 is a conceptual diagram for explaining topic models and data collected by the related data collection unit 106, and their relationships. FIG. 14 shows a screen 1401 in which topic models and data collected by the related data collection unit 106 and their relationships are extracted and displayed from among the topic models and data shown in the conceptual diagram shown in FIG. ing. The related data display control unit 107 displays such information on the screen of the display unit 104.

In FIG. 14, the related data collection unit 106 collects shipment trend data 30211 by machine, coal production trend data 3022, power generation data 3033 by country and energy as the above-mentioned source data of the external topic model 1112, and sales data 2023, Customer data 2024 and product data 2026 are collected as the source data of the user topic model 1111, and the related data display control unit 107 generates visualization data and keywords that visualize the collected data and the topic model containing the data. Display data (provided data) is generated and displayed on the screen of the display unit 104. Further, there is a relationship 1301 between the coal production amount transition data 3022 and the customer data 2024, and a relationship 1302 exists between the machine-by-machine shipment transition data 30211 and the product data 2026. Relevance 1301 and Relevance 1302 refer to the feedback repository 109 shown in FIG. 7, refer to the external data topic name and feedback topic name of the newly added record, and create a topic model containing the above source data. It may be determined that there is a relationship between them. Although FIG. 14 illustrates a case where the original data are related to each other, the same can be considered between topic models. Further, in FIG. 14, the related data display control unit 107 displays keywords such as "coal" and "bulldozer."

Thereafter, the data transmitting unit 108 transmits the provision data generated by the related data display control unit 107 to the user server 200 as problem analysis solution data 2031. A person in charge of the user company receiving consulting operates an input device (for example, a keyboard) included in the server 200 to transmit the provision data received by the server 200 on a display device such as a display included in the server 200. It is displayed on a certain input display section 203 and its contents are confirmed.

In this way, the feedback repository 109 shown in FIG. 7 stores topic names of external topic models that are related to each other and feedback topic names that are topic names of user topic models in association with each other. Display gender on screen. Therefore, the relationship between the user data and the externally published data, which cannot be obtained only by analysis using the externally published data 300 and the user topic model 1111 based on the user data including user private data, is User companies can present new insights without having to be aware of the differences in terms used between them.

FIG. 9 is a flowchart showing detailed steps of the processing in S812 and S820 shown in FIG. As shown in FIG. 9, the analysis setting unit 103 selects data designated by the analyst for use in analysis from the external public data 300 copied to the consulting server 100 (S901). The analysis setting unit 103 determines whether or not to select an external factor as an axis of analysis from the feedback repository 109 (S902), and if it is determined that an external factor as an axis of analysis is to be selected from the feedback repository 109 (S902; Yes), reads the external topic name and domain of the feedback repository 109 that is being generated and updated at that time, and uses the topic model corresponding to the read external topic name and domain and the data that is the source of the topic model as an external factor. Select (S903). When the topic model to be used as an external factor and the source data of the topic model are selected in S903, the process advances to S814. In S814, an analysis is performed using a topic model of the external topic name selected as an external factor and the data that is the source of the topic model, using a predetermined analysis method and a threshold used in the analysis method. .

FIG. 10 is a flowchart showing the detailed procedure of the process of S824 shown in FIG. As shown in FIG. 10, the related data collection unit 106 collects the topic model used to output the item 1204 for which the alert was output as an analysis result and the data that is the source of the topic model to the topic model analysis unit 111. (S1001). The related data collection unit 106 determines whether or not to utilize the feedback repository 109 based on the determination result in S813 (S1002), and if it is determined to utilize the feedback repository 109 (S1002; Yes), the external topic of the feedback repository 109 is The external topic name and domain are read out, and the topic model corresponding to the read external topic name and domain and the source data of the topic model are collected as related data (S1003). When items to be related data are collected in S1003, the process advances to S825.

FIG. 11 is a flowchart showing detailed steps of the processing from S815 to S819 shown in FIG. As shown in FIG. 11, first, the topic model analysis unit 111 uses various conventionally known cluster analyzes to identify items whose distance between items is equal to or less than a predetermined threshold in each topic model. A topic name is specified (S1101).

The topic model analysis unit 111 determines whether the distance between the topic name of the mixed topic model and the keyword that is the origin of the topic name is less than or equal to the threshold, among the topics that include items for which the distance between the items is less than or equal to a predetermined threshold. The topic name of the user topic model 1111 corresponding to the topic name is extracted (S1102). For example, the topic name of the user topic model 1111 whose distance is zero is extracted as a topic name synonymous with the keyword.

In addition, the topic model analysis unit 111 determines that, among topics that include items for which the distance between the items is equal to or less than a predetermined threshold, the distance between the topic name of the mixed topic model and the keyword that is the source of the topic name is The topic name of the external topic model 1112 corresponding to the topic name that is equal to or less than the threshold value and the topic name of the external topic model 1112 that is different from the topic name are extracted (S1103). For example, the topic name of the external topic model 1112 whose distance is zero is extracted as a topic name synonymous with the keyword, and a topic name that is semantically closer than a certain level to the topic name is extracted as the different topic name.

The feedback learning unit 110 associates the topic name of the user topic model 1111 extracted in S1102 with the topic name of the external topic model 1112 extracted in S1103 and the above-mentioned different topic names as mutually related topics, and provides feedback. It is added to the repository 109 as an alias record (S1104).

After that, the feedback learning unit 110 sets the name of the business type or industry to which the user company included in the IR data 2011 belongs to the domain of the added record (S1105).

As described above, according to the consulting support system 1000 of this embodiment, when performing analysis using a topic model, it is possible to resolve differences in definitions of words in domains. For example, by maintaining a feedback repository as shown in Figure 7 on the system and associating external topic models with user topic models, it is possible to perform analysis that takes into account differences in terms in external topic models and externally published data. and present the results.

Furthermore, as shown in Figure 12, if the analysis result is not good, that item is made selectable, and when the item is selected, the topic model that was the basis for that item and the topic By making the data that is the basis of the model accessible, it is possible to recommend the data necessary to solve problems in items analyzed as unfavorable.

Note that in the above embodiment, in FIG. 8, the processes from S815 to S819 for generating and updating the feedback repository 109 were described as one analysis process. However, the processes from S815 to S819 may be executed as one feedback repository generation process independently from other processes.

1000 Consulting support system 100 Consulting server 200 User servers N1, N2 Network 101 User data acquisition section 102 External data acquisition section 103 Analysis setting section 104 Display section 105 Alert processing section 106 Related data collection section 107 Related data display control section 108 Data transmission section 109 Feedback repository 110 Feedback learning section 111 Topic model analysis section 112 Topic model generation section 201 Public data storage section 202 Internal data storage section 203 Input display section 204 Display section

Claims (6)

A user topic model obtained from user data including user public data and user private data owned by the user company, and an external topic model obtained from external public data published outside of the user company. , the semantic similarity between the topic name of the user topic model and the topic name of the external topic model, which is obtained using a predetermined method from the user data and the externally published data, is calculated between the items of the topic model. a topic model generation unit that generates a mixed topic model that is a topic model common to the external topic model and the user topic model, for making a determination based on the distance of ;
A feedback topic that is a topic in which the distance between a topic included in the mixed topic model and a topic included in the user topic model is equal to or less than a threshold is identified from the user topic model, and a feedback topic is identified from the user topic model, and a topic included in the mixed topic model and the external topic are identified. a topic model analysis unit that identifies an external topic from the external topic model, which is a topic whose distance from the topic of the model is equal to or less than a threshold;
a feedback learning unit that associates the identified feedback topic with the external topic and registers it in a feedback repository;
A consulting support system characterized by having. an analysis setting unit that defines items that serve as analysis indicators for the user data as internal factors, and defines items that serve as analysis indicators for the externally disclosed data as external factors;
The analysis setting unit reads the registered feedback repository, redefines items that serve as analysis indicators for the defined externally published data, and performs a process of analyzing the user company using a predetermined analysis method.
The consulting support system according to claim 1, characterized in that: The topic model analysis unit selects a topic name for which a distance between a topic name of the mixed topic model or a keyword that is a source word of the topic name and a topic corresponding to the topic name of the user topic model is less than or equal to a threshold value. The topic name of the corresponding user topic model is extracted as the feedback topic , and the distance between the topic name or the keyword and the topic corresponding to the topic name of the external topic model is equal to or less than a threshold value. Extract the topic name of the external topic model and the topic name of the external topic model different from the topic name,
The feedback learning unit associates the topic name of the user topic model, the topic name of the external topic model, and the topic name of the different external topic model as mutually related topics, and adds them to the feedback repository.
The consulting support system according to claim 1, characterized in that: The analysis setting section displays the results analyzed by the predetermined analysis method on a display section,
an alert processing unit that notifies an alert when the analyzed result does not satisfy a threshold value and makes the corresponding item in the analyzed result selectable on the display unit;
3. The consulting support system according to claim 2, further comprising: When the item in the selectable state is selected, the topic used to output the item for which the alert was output as the result of the analysis as related data that is the basis for the selected item. A related data collection unit that collects the data that is the basis of the model and the topic model;
5. The consulting support system according to claim 4, further comprising: A consulting support method for supporting consulting for a user company using a computer,
A user topic model obtained from user data including user public data and user private data owned by the user company, and an external topic model obtained from external public data published outside of the user company. , the semantic similarity between the topic name of the user topic model and the topic name of the external topic model, which is obtained using a predetermined method from the user data and the externally published data, is calculated between the items of the topic model. generating a mixed topic model that is a topic model common to the external topic model and the user topic model, for making a determination based on the distance of ;
Identifying a feedback topic from the user topic model that is a topic in which a distance between a topic included in the mixed topic model and a topic included in the user topic model is equal to or less than a threshold;
Identifying an external topic from the external topic model that is a topic in which a distance between a topic included in the mixed topic model and a topic included in the external topic model is equal to or less than a threshold;
associating the identified feedback topic with the external topic and registering it in a feedback repository;
A consulting support method characterized by:

Images