JP7471760B1 - Information processing method, information processing system, and program - Google Patents

Abstract

[Problem] To provide an information processing method, information processing system, and program in which a large-scale language model outputs evaluation information. [Solution] The information processing method includes acquiring public information of a subject, generating instruction information requesting output of evaluation information that evaluates the subject from a predetermined viewpoint based on the public information, inputting the instruction information into the predetermined large-scale language model, and the predetermined large-scale language model outputting the evaluation information. [Selected Figure] Figure 3

Description

The present invention relates to an information processing method, an information processing system, and a program.

Patent document 1 discloses a tool that allows easy analysis of segments for appropriate marketing activities.

Japanese Patent No. 6704163

The technology in Patent Document 1 divides survey respondents into segments based on survey information. However, because it is based on survey information, there is room for improvement in terms of accuracy.

According to one aspect of the present invention, an information processing method executed by an information processing system is provided. This information processing method acquires public information of a subject. It generates instruction information requesting output of evaluation information that evaluates the subject from a predetermined viewpoint based on the public information. It inputs the instruction information into a predetermined large-scale language model. The predetermined large-scale language model outputs the evaluation information.

It is possible to obtain information that can serve as an appropriate indicator to replace survey information.

FIG. 1 is a diagram illustrating an example of a system configuration of an information processing system. FIG. 2 is a diagram illustrating an example of a hardware configuration of an information processing device. FIG. 3 is a sequence diagram illustrating an example of information processing in the information processing system. FIG. 4 is a diagram showing an example of a list of potential customers.

The following describes embodiments of the present invention with reference to the drawings. The various features shown in the following embodiments (including modified examples, which apply hereinafter) can be combined with each other.

<Embodiment 1>
1. System Configuration Diagram Fig. 1 is a diagram showing an example of the system configuration of an information processing system 1000. As shown in Fig. 1, the information processing system 1000 includes, as a system configuration, an information processing device 100 and a system that provides a function of a large-scale language model (hereinafter simply referred to as a large-scale language model) 110. The information processing device 100 and the large-scale language model 110 are communicatively connected via a network 150. The information processing device 100 is an example of a computer.

The information processing device 100 executes processing related to the first embodiment. Details of the processing of the information processing device 100 will be described later with reference to FIG. 3 and the like. The large-scale language model 110 is a natural language processing model trained using a large amount of text data. The large-scale language model 110 of the first embodiment is preferably a large-scale language model tuned particularly for chat (dialogue).

Here, the information processing system described in the claims may be composed of multiple devices, or may be composed of one device. When the information processing system described in the claims is composed of one device, an example of that device is the information processing device 100. When the information processing system described in the claims is composed of multiple devices, an example of the multiple devices is the information processing device 100 and the large-scale language model 110, etc.

2. Hardware Configuration FIG. 2 is a diagram showing an example of the hardware configuration of the information processing device 100.
As shown in FIG. 2, the information processing device 100 includes, as its hardware configuration, a control unit 210, a storage unit 220, an input unit 230, an output unit 240, and a communication unit 250.

The control unit 210 is a CPU (Central Processing Unit) or the like, and controls the entire information processing device 100. The storage unit 220 is any one of an HDD (Hard Disk Drive), a ROM (Read Only Memory), a RAM (Random Access Memory), an SSD (Solid State Drive), or any combination thereof, and stores a program and data (e.g., the acquired public information, a potential customer list, etc., described below) used when the control unit 210 executes processing based on the program. The storage unit 220 is an example of a storage medium. In the specification, the data used when the control unit 210 executes processing based on the program is described as being stored in the storage unit 220, but it may be stored in a storage unit of another device that can communicate with the information processing device 100. In other words, the data may be stored in the storage unit of any device as long as the control unit 210 can refer to it. The control unit 210 executes processing based on the program stored in the storage unit 220, thereby realizing the functions of the information processing device 100 and the processing of the information processing device 100 in the sequence diagram shown in FIG. 3, which will be described later.

The input unit 230 is a keyboard, a mouse, or the like, and inputs information based on a selection operation and/or an input operation by an operator.
The output unit 240 is a display or the like, and displays information input by an operator via a screen, results of processing by the control unit 210, and the like.
The communication unit 250 connects the information processing device 100 to the network 150 and controls communication with other devices.

3. Information Processing Hereinafter, information processing in the first embodiment will be described.

(1) Overview of Processing The control unit 210 of the information processing device 100 acquires public information of the subject. Here, the public information is information that the subject (company and/or individual) has made public, and is either or both of document or text information related to the subject and image information related to the subject. When the subject is a company, examples of the public information include document or text information in the company's web page and either or both of image information in the company's web page. The web page is one example, and other examples include a securities report, a financial statement, and the like. In addition, the image information related to the subject may include a capture image of the web page related to the subject. That is, the capture image of the company's web page may also be included in the image information related to the company that is the subject. The capture image of the web page may be a capture image of the entire web page, or a capture image of a part of the web page. When the subject is an individual, the public information may be either or both of document or text information included in information transmitted by the individual through a personal web page, a paper, or a publicly-available SNS (Social Networking Service), and image information. When the subject is an individual, a captured image of the personal web page may also be included in the image information related to the individual who is the subject. The control unit 210 generates a prompt. Here, the prompt is a prompt including content requesting output of evaluation information that evaluates the subject based on the public information from a predetermined viewpoint. The prompt is an example of instruction information. The control unit 210 inputs the generated prompt to the large-scale language model 110. The large-scale language model 110 is an example of a predetermined large-scale language model. The large-scale language model 110 outputs the evaluation information.
By such processing, it is possible to obtain information that can serve as an appropriate indicator to replace questionnaire information, that is, evaluation information.
(2) Details of Processing Fig. 3 is a sequence diagram showing an example of information processing in the information processing system 1000. In the example of Fig. 3, the information processing in the information processing system 1000 will be described by taking as an example a process of generating a list of potential customers for a product to be sold or a service to be provided. The product to be sold or the service to be provided are examples of a predetermined product or service. For the sake of simplicity, the following description will be given by taking the product to be sold as an example.
Upon receiving predetermined operation information, in sequence SQ301, the control unit 210 determines the target person based on attribute information. For example, the control unit 210 determines the target person based on attribute information of existing customers of the product being sold. Here, the attribute information of existing customers is information that indicates the attributes of existing customers. For the sake of simplicity, let us take a company as an example of the target person. Examples of attribute information of existing customers include company size, sales, number of employees, etc. The control unit 210 determines the target person to be a company similar to this information based on the company size, sales, number of employees, etc. of existing customers of the product being sold.

In sequence SQ302, the control unit 210 acquires public information about the target determined in sequence SQ301. As described above, when the target is a company, examples of public information include the company's web page, securities report, financial statements, etc. For example, the control unit 210 uses web scraping technology to acquire information such as the target's web page, securities report published on the web, and financial statements published on the web.

In sequence SQ303, the control unit 210 determines a predetermined perspective based on information about the product being sold. Information about the product being sold includes, for example, product introduction materials, product sales materials, product pamphlets, etc. For example, the control unit 210 recognizes the characters written in the product introduction materials, product sales materials, product pamphlets, etc., extracts the product's features, and determines a predetermined perspective based on the product's features. Examples of the predetermined perspective include, for example, "easy-to-work workplace," "promotes DX (Digital Transformation)," and "enthusiastic about employee training." There may be one or more predetermined perspectives.

As another example of determining a specified perspective, the control unit 210 may determine a perspective using a trained model that has been trained using information about the product being sold, such as product introduction materials, sales materials, and product brochures, as input data, and a perspective corresponding to the product as output data. That is, the control unit 210 inputs information about the product being sold, such as product introduction materials, sales materials, and product brochures, into the trained model. The control unit 210 may then determine the perspective output from the trained model as the perspective of the product being sold. The same applies to the case of a service if product is read as service.

The processing of sequence SQ303 may be performed simultaneously with sequence SQ301 or sequence SQ302, or may be performed before sequence SQ301 or sequence SQ302.

In sequence SQ304, the control unit 210 generates a prompt based on the viewpoint determined in sequence SQ303, the subject selected in sequence SQ301, and the public information acquired in sequence SQ302. More specifically, the control unit 210 generates a prompt requesting the output of evaluation information that evaluates the subject selected in sequence SQ301 from the viewpoint determined in sequence SQ303 based on the public information acquired in sequence SQ302. The control unit 210 may include information specifying the output format of the evaluation information in the prompt. This is because it is more convenient to use the output evaluation information if the output format is a predetermined output format. The control unit 210 may also include examples of input and output in the prompt. This is to improve the accuracy of the output. In addition, if the public information is text information, the control unit 210 may include information instructing the subject to be evaluated for each text information of a predetermined number of characters in the prompt, and if the public information is image information, the control unit 210 may include information instructing the subject to be evaluated for each image information (for example, an image included in a web page) in the prompt.

In sequence SQ305, the control unit 210 inputs the prompt generated in sequence SQ304 to the large-scale language model 110. For example, the control unit 210 inputs the prompt to the large-scale language model 110 using an API (Application Programming Interface).

In sequence SQ306, the control unit 210 receives the evaluation information output by the large-scale language model 110 via the API. The evaluation information may be, for example, information indicating a score out of 100 for each predetermined perspective. For example, if the perspectives are "easy-to-work-in workplace," "promoting DX," and "enthusiastic about employee training," the control unit 210 receives the evaluation information as "easy-to-work-in workplace" 80 points, "promoting DX" 65 points, and "enthusiastic about employee training," etc.

In sequence SQ307, the control unit 210 determines whether the subject satisfies the predetermined requirements based on the received evaluation information. If the control unit 210 determines that the subject satisfies the predetermined requirements based on the evaluation information, it advances the process to sequence SQ308. If the control unit 210 determines that the subject does not satisfy the predetermined requirements based on the evaluation information, it ends the information processing. For example, if the predetermined requirement is "70 points or more in all respects," and the evaluation information received from the large-scale language model 110 is "easy-to-work workplace" 80 points, "promoting DX" 65 points, "enthusiastic about employee training" 85 points, etc., the control unit 210 determines that the subject does not satisfy the predetermined requirements. If the predetermined requirement is "70 points or more in all respects," and the evaluation information received from the large-scale language model 110 is "easy-to-work workplace" 80 points, "promoting DX" 95 points, "enthusiastic about employee training" 85 points, etc., the control unit 210 determines that the subject satisfies the predetermined requirements.

In sequence SQ308, the control unit 210 adds the target person to a list of potential customers for the product being sold.
Fig. 4 is a diagram showing an example of a list of potential customers. List 400 is a list of potential customers. The list includes information on targets as potential customers. In the example of Fig. 4, company names, addresses, telephone numbers, and email addresses are included as targets.
For example, when the control unit 210 finishes creating the list 400, it sends sales emails or the like about the products being sold to the targets included in the list 400.

According to the processing of the first embodiment, it is possible to obtain information that serves as an appropriate indicator to replace questionnaire information, i.e., evaluation information. Furthermore, if the evaluation information meets certain requirements, the target person can be added to the list. In the example of the embodiment described above, a list of potential customers for a product can be created quickly and based on appropriate indicators.

Note that, in the first embodiment, an example of creating a list of potential customers has been described, but the present invention is not limited to this. For example, if the subjects are applicants for employment at a company, and the predetermined viewpoint is qualitative information indicating the company's values derived from the company's website or brochures, the control unit 210 can generate instruction information requesting the output of evaluation information that evaluates the applicant based on the applicant's public information and information indicating the company's values, etc., and input this to the large-scale language model. The control unit 210 can then receive the applicant's evaluation information from the large-scale language model.

(Variation 1)
Modification 1 of the first embodiment will be described. Note that the modification is included in the first embodiment and does not indicate other embodiments. The same applies to the following modifications. Modification 1 will be mainly described with respect to the points that are different from the first embodiment.
In the first embodiment, the large-scale language model 110 is provided as an entity separate from the information processing device 100. However, the large-scale language model 110 may be implemented in the information processing device 100.
The configuration of the first modified example can achieve the same effects as those of the first embodiment.

(Variation 2)
Modification 2 of embodiment 1 will be described below. In modification 2, differences from embodiment 1 will be mainly described.
In the first embodiment, a large-scale language model has been described as an example. However, instead of the large-scale language model, a trained model that is trained specifically for the purpose may be used.
The configuration of the second modification can achieve the same effects as those of the first embodiment.

(Variation 3)
Modification 3 of embodiment 1 will be described below. In modification 3, differences from embodiment 1 will be mainly described.
In the first embodiment, a series of processes including inputting a prompt, obtaining an output result, making a judgment, and adding a candidate to the list are performed. The control unit 210 of the third modification may perform this series of processes multiple times and update the list of candidates based on the results. Furthermore, the control unit 210 of the third modification may obtain an output result by paraphrasing the prompt each time the series of processes are performed multiple times.
The configuration of variant example 3 achieves the same effects as embodiment 1, and can obtain a list of highly accurate candidates even when the output from the trained model is unstable.

Finally, although various embodiments of the present invention have been described, these are presented as examples and are not intended to limit the scope of the invention. New embodiments can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. The embodiments and their modifications are within the scope and spirit of the invention, and are also within the scope of the invention and its equivalents as set forth in the claims.

100: Information processing device 110: Large-scale language model 150: Network 210: Control unit 220: Storage unit 230: Input unit 240: Output unit 250: Communication unit 400: List 1000: Information processing system

Claims (10)

An information processing method executed by an information processing system, comprising:
determining a viewpoint of the product or service to be sold based on information about the product or service to be sold;
The viewpoint is a qualitative viewpoint,
Obtain public information about the subject;
generating instruction information for requesting output of evaluation information for evaluating the subject based on the public information from the perspective of the determined product or service ;
inputting said instructional information into a predetermined large-scale language model;
The predetermined large-scale language model outputs the evaluation information;
The evaluation information is information in which a quantitative evaluation value is assigned to the qualitative viewpoint.
Information processing methods. 2. The information processing method according to claim 1,
If the quantitative evaluation value is equal to or greater than a threshold, the target person is added to a list of customer candidates for the product or service to be sold.
Information processing methods. 2. The information processing method according to claim 1,
determining the target person based on attribute information;
Obtaining the public information of the determined subject;
Information processing methods. 2. The information processing method according to claim 1,
determining said target persons based on attribute information of existing customers of the product or service to be sold ;
Obtaining the public information of the determined subject;
Information processing methods. 5. The information processing method according to claim 4 ,
determining targets similar to the attribute information of the existing customers based on the attribute information of the existing customers of the product or service to be sold;
Information processing methods. 2. The information processing method according to claim 1,
The public information is either or both of publicly available document information related to the subject and publicly available image information related to the subject;
Information processing methods. 7. The information processing method according to claim 6,
the image information relating to the subject includes a captured image of a web page relating to the subject;
Information processing methods. 2. The information processing method according to claim 1,
The instruction information includes information for specifying an output format of the evaluation information.
Information processing methods. An information processing system,
At least one control unit is included,
The control unit is
determining a viewpoint of the product or service to be sold based on information about the product or service to be sold;
The viewpoint is a qualitative viewpoint,
Obtain public information about the subject;
generating instruction information for requesting output of evaluation information for evaluating the subject based on the public information from the perspective of the determined product or service ;
inputting said instructional information into a predetermined large-scale language model;
The predetermined large-scale language model outputs the evaluation information;
The evaluation information is information in which a quantitative evaluation value is assigned to the qualitative viewpoint.
Information processing system. A program,
On the computer,
A program for executing the information processing method according to any one of claims 1 to 8.