Marketing activity support device

Abstract

To provide a marketing activity support device for sufficiently supporting marketing activities by generating a realistic role playing model based on a script.SOLUTION: A marketing activity support device 1 includes: a talk script extraction unit 112 for extracting talk scripts that have succeeded in marketing activities, on the basis of new target information, and target information and success / failure information of a marketing database 121; a large-scale language model unit 113 having a large-scale language model specialized for creating hypothetical targets and role-playing models, and a role playing model generation unit 114 that generates a hypothetical target on the basis of the new target information by outputting a prompt to the large-scale language model unit 113 and generates a role-playing model in a case where the talk script extracted by the talk script extraction unit 112 is applied to the hypothetical target.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a marketing activity support device. [Background technology]

[0002] In the field of sales strategies, attempts to improve sales efficiency by collecting and analyzing information about customers or potential customers are actively underway.

[0003] For example, Patent Document 1 discloses a technology that includes a sales data acquisition means that acquires information accumulated when a sales entity conducts sales activities as sales data, a sales data analysis means that analyzes the sales data, and a target list generation means that generates a target list listing candidates for targets that the sales entity will call based on the analysis results from the sales data analysis means, and that generates a talk script that indicates the flow of conversation that an operator should have when calling a target based on the analysis results, and that applies a machine learning method to information that includes at least the newly acquired sales data, and updates the algorithm used in list generation by the target list generation means based on the results.

[0004] As a result, the technology of Patent Document 1 makes it possible to efficiently carry out sales activities by collecting and organizing a large amount of information. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Publication No. 2019-164648 Summary of the Invention [Problem to be solved by the invention]

[0006] Incidentally, the technology of Patent Document 1 makes it possible to support marketing activities by generating talk scripts, but there is room for further improvement in that it is unclear how the generated talk scripts can be used in marketing activities aimed at targets.

[0007] The present invention has been made in light of the above circumstances, and its purpose is to provide a marketing activity support device that can fully support marketing activities by generating a realistic role-playing model based on a script. [Means for solving the problem]

[0008] As a result of extensive research into solving the above-mentioned problems, the inventors have discovered that the above-mentioned object can be achieved by receiving new target information, extracting an optimal talk script from past successful examples, and using a large-scale language model specialized in creating hypothetical targets and role-playing models to generate a realistic role-playing model based on the talk script. The inventors have then completed the present invention. Specifically, the present invention provides the following:

[0009] The present invention provides a marketing database in which target information consisting of a plurality of target elements relating to a target in a marketing activity, a talk script for the target, and success / failure information indicating the success or failure of the marketing activity according to the talk script are stored in association with each other; an information receiving unit that receives new target information consisting of one or more target elements; a talk script extraction unit that extracts the talk script that made the marketing activity successful based on the new target information, the target information in the marketing database, and the success / failure information; a large-scale language model unit that combines one or more of the natural language processing methods based on the target information in the marketing database to probabilistically predict how likely words and sentences given in a prompt are to occur in natural language, and that performs learning and adjustment to generate a hypothetical target and a role-playing model through sentence generation and question answering; and a role-playing model generation unit that generates the assumed target based on the new target information by outputting the prompt to the large-scale language model unit, and generates the role-playing model when the talk script extracted by the talk script extraction unit is applied to the assumed target; The marketing activity support device has the above.

[0010] According to the present invention, it is possible to fully support marketing activities by receiving new target information, extracting the optimal talk script from past successes, and using a large-scale language model specialized in creating hypothetical targets and role-playing models to generate a realistic role-playing model based on the talk script. [Effects of the Invention]

[0011] According to the present invention, a realistic role-playing model is generated based on a script, thereby making it possible to fully support marketing activities. [Brief explanation of the drawings]

[0012] [Figure 1] FIG. 1 is an explanatory diagram showing the flow of information in the marketing activity support device of this embodiment. [Figure 2] FIG. 2 is an explanatory diagram showing the flow of information in the role-playing model creating device of this embodiment. [Figure 3]FIG. 3 is a block diagram showing an example of the hardware configuration of the role-playing model creating device of this embodiment. [Figure 4] FIG. 4 is an explanatory diagram showing a display screen of a user terminal. [Figure 5] FIG. 5 is an explanatory diagram showing the flow of information in the marketing activity support device of this embodiment. [Figure 6] FIG. 6 is an explanatory diagram of the marketing database. [Figure 7] FIG. 7 is an explanatory diagram of the marketing database. [Figure 8] FIG. 8 is an explanatory diagram of a marketing activity support device that executes a marketing activity program. DETAILED DESCRIPTION OF THE INVENTION

[0013] Hereinafter, an example of an embodiment of the present invention will be described in detail with reference to the drawings. (Marketing Activity Support Device 1) As shown in Figure 1, the marketing activity support device 1 is configured to receive new target information, extract the optimal talk script from past success stories, and generate a realistic role-playing model based on the talk script using a large-scale language model specialized for creating hypothetical targets and role-playing models.

[0014] To give a specific example, the marketing activity support device 1 has a marketing database 121 (also simply referred to as "database 121") in which target information consisting of multiple target elements related to targets in marketing activities, talk scripts for the targets, and success / failure information indicating the success or failure of marketing activities using the talk scripts are stored in association with each other, an information receiving unit 111, and a talk script extraction unit 112.

[0015] Here, "marketing activities" refers to all sales promotion activities for products and services that a company undertakes to customers. Specific examples include market research, advertising, sales promotion, customer relationship management, product planning and development, pricing, distribution and sales, and customer service. "Target" refers to the specific customers or customer segments considered most likely to purchase products or services in marketing activities. For example, targets can be set based on attributes such as age, gender, occupation, income, interests, lifestyle, values, needs, challenges, and region.

[0016] "Target elements" refer to the elements that make up a target. Examples include name, age, gender, occupation, address, phone number, email address, interests, lifestyle, values, needs, challenges, purchase history, and customer satisfaction. Specifically, these elements include age: 20s to 30s, gender: female, occupation: office worker, income: annual income of 5 million yen or more, interests: beauty, fashion, lifestyle: urban living, values: nature-oriented, needs: skin without blemishes or wrinkles, and challenges: dry skin. "Target information" is a collection of multiple target elements.

[0017] "Talk scripts" are examples of conversations that salespeople use when meeting with customers. Specifically, talk scripts include content such as greetings, self-introductions, confirming customer needs, proposing solutions, handling objections, closing, and follow-up. "Success / failure information" is information that indicates the success or failure of marketing activities. Specifically, this information includes sales, profits, number of customers acquired, customer satisfaction, and customer dropout rate.

[0018] It is preferable that the target information, talk script, and success / failure information be stored in the marketing database 121 in a state normalized into a unified data format. In this case, when handling large amounts of data, the data is converted into a common format and uniformly organized, so that when learning, the normalized information can be used as input to improve processing efficiency and accuracy. "Normalization" is a processing technique for organizing and structuring data in database tables according to certain rules, thereby eliminating data redundancy and maintaining data consistency and integrity.

[0019] The information receiving unit 111 has a function of receiving new target information consisting of one or more target elements. Here, "new target information" is information about a target that is input by a sales entity, i.e., a user 4, into the user terminal 2 and received by the information receiving unit 111. The "sales entity" is a company or organization, a sole proprietor, a sales representative, etc., that conducts sales activities.

[0020] The information receiving unit 111 also has an activation function. The activation function enables data communication only with authenticated user terminals 2 and administrator terminals 7. The administrator terminal 7 is a terminal device operated by a person in charge of managing the marketing activity support device 1. Specifically, the activation function verifies whether an activation signal transmitted from a specific user terminal 2 or the like matches pre-set authentication information. Examples of authentication methods include authentication modes that combine one or more authentication elements, such as password authentication, biometric authentication, one-time password authentication, and smart card authentication using a card with an embedded IC chip. This allows the marketing activity support device 1 to communicate data only with authenticated user terminals 2 and administrator terminals 7, making it difficult for information to leak.

[0021] The activation function may also be provided between control unit 11 and memory unit 12 having marketing database 121, allowing data communication only between authenticated control unit 11 and marketing database 121. In this case, if memory unit 12 is installed in a location away from control unit 11 and is accessible via data communication such as the Internet 5, it is possible to prevent unauthorized access to marketing database 121 from outside.

[0022] The talk script extraction unit 112 has a function of extracting talk scripts that have made marketing activities successful, based on new target information and target information and success / failure information in the marketing database 121. Examples of methods for extracting talk scripts include a rule-based method, a template matching method, a heuristic approach method based on empirical rules, and a clustering method.

[0023] The rule-based method extracts talk scripts based on predefined rules and patterns. For example, it extracts sentences containing specific keywords or phrases as talk scripts. The template matching method extracts appropriate talk scripts by matching them with pre-defined talk script templates. By selecting templates with high similarity, talk scripts that suit the target are extracted. The heuristic approach extracts talk scripts using statistical techniques or heuristic algorithms. For example, it selects talk scripts by analyzing the frequency and patterns of talk scripts that meet specific conditions. The clustering method clusters talk scripts with similar characteristics and extracts representative talk scripts from each cluster. Grouping talk scripts based on similarity enables effective extraction.

[0024] In addition, the talk script extraction unit 112 may be configured to use target information, talk scripts, and suitability information as training data, and extract talk scripts that are likely to lead to successful marketing activities for new target information through machine learning equipped with a talk script prediction model trained to predict successful talk scripts based on the target information and success / failure information.

[0025] Here, the "talk script prediction model" is a machine learning model designed to predict the most effective talk script for a new target based on past data from marketing activities. This talk script prediction model uses existing target information, talk scripts, and success / failure information as training data and learns patterns from this data. Through the learning process, the talk script prediction model understands which types of talk scripts have brought success in specific target information and scenarios, and acquires the ability to select talk scripts with a high probability of success when new target information is given.

[0026] According to the above configuration, by using a talk script prediction model trained through machine learning, it is possible to improve the probability of success in marketing activities targeting new targets, and by extracting optimal talk scripts from past success stories, it is possible to achieve an effective approach to new targets.

[0027] The marketing activity support device 1 further includes a large-scale language model unit 113 and a role-playing model generation unit 114. The large-scale language model unit 113 combines one or more types of natural language processing based on the target information in the marketing database 121 to probabilistically predict how likely words and sentences given in the prompt are to occur in natural language, and performs learning and adjustment to generate a hypothetical target and a role-playing model through sentence generation and question answering, thereby providing a large-scale language model specialized for creating a hypothetical target and a role-playing model. The role-playing model generation unit 114 outputs a prompt to the large-scale language model unit 113 to generate a hypothetical target based on new target information and generate a role-playing model when the talk script extracted by the talk script extraction unit 112 is applied to the hypothetical target.

[0028] Here, "natural language processing" refers to a process that enables a computer to understand text and audio data written in natural language and execute processing appropriate to the purpose. Specifically, examples include morphological analysis, which breaks natural language down into "morphemes," the smallest units that make up the language, and assigns information such as parts of speech; syntactic analysis, which analyzes the grammatical structure of natural language to clarify the structure and meaning of a sentence; semantic analysis, which analyzes the meaning of natural language to understand the meaning of words and sentences and make logical judgments and inferences; contextual analysis, which understands natural language while taking into account the context before and after a sentence; and intent analysis, which extracts the intention of a speaker or writer from a conversation or sentence using natural language. Thus, "natural language processing" processes natural language by combining processes such as morphological analysis, syntactic analysis, semantic analysis, contextual analysis, and intent analysis, and enables the generation of role-playing models, machine translation, automatic summarization, question-answering systems, and speech recognition that support the marketing activities of this embodiment.

[0029] A "prompt" is a word or sentence input to the large-scale language model unit 113, and serves as a starting point for the large-scale language model to generate a hypothetical target and a role-playing model. In other words, the prompt is an important input for the large-scale language model to generate content tailored to a specific target demographic or marketing scenario. Specifically, the prompt includes keywords and phrases related to new target information or a specific marketing scenario, and based on these inputs, the large-scale language model generates a talk script appropriate for the target demographic and a role-playing model to be used by the sales entity.

[0030] The generation of role-playing models using a large-scale language model based on prompts is a major difference from regular (conventional) machine learning. In regular machine learning, a model learns from training data and predicts output data for input data. For example, a machine learning model for image recognition learns from training data of images of cats and dogs and classifies the input image as either a cat or a dog. On the other hand, in role-playing model generation using a large-scale language model, the model learns not only from training data but also from prompts that provide instructions and information about the output data that the model should generate.

[0031] In more detail, while conventional machine learning models can only generate content contained in the training data, large-scale language models can generate new content not contained in the training data through prompts. For example, even if the training data only contains the conversations of female office workers in their 20s, prompts can also generate the conversations of male office workers in their 50s. Furthermore, while conventional machine learning models can only generate variations of the content contained in the training data, large-scale language models can generate new variations of output data not contained in the training data through prompts. For example, even if the training data only contains conversations from the perspective of sales representatives, prompts can also generate conversations from the perspective of customers. Furthermore, while conventional machine learning models can only generate new content by recursively combining content contained in the training data, large-scale language models can generate creative content not contained in the training data through prompts. For example, even if the training data only contains descriptions of existing products and services, prompts can generate ideas for completely new products and services.

[0032] The "hypothetical target" is a profile of an assumed target demographic generated based on new target information and existing target information. This hypothetical target is generated by the large-scale language model unit 113 based on the target information and success / failure information stored in the marketing database 121, and is used to increase the probability of success in marketing activities. For example, if information such as "age: 30s, interests: health, lifestyle: active" is input as new target information, the large-scale language model unit 113 creates a hypothetical target based on this information, referring to past success stories and related target information, to generate a talk script appropriate for this target demographic. For example, the hypothetical target is expressed as a specific target profile (e.g., "people in their 30s with an active lifestyle who are highly interested in health").

[0033] A "role-playing model" refers to a model used by a sales entity to simulate a conversation with a target customer. The role-playing model is generated by combining a hypothetical target and a talk script extracted by the talk script extraction unit 112. Specifically, the role-playing model anticipates responses and questions based on the hypothetical target and presents a talk script for the sales entity to use. This enables the sales entity to communicate more appropriately in conversations with actual targets. For example, if the hypothetical target is "health-conscious people in their 30s with active lifestyles," the role-playing model suggests ways to introduce products and services that this target demographic might be interested in, answers to potential questions they might have, and methods for handling objections. This allows the role-playing model to provide practice for the sales entity to gain confidence and communicate effectively in actual conversations with targets. This can improve the results of marketing activities.

[0034] The role-playing model includes existing list appointment-making talks, such as an elevator pitch talk for making an appointment and automatic generation of appointment-making talk scripts. Here, "existing list appointment-making talks" are a list of sales talks for obtaining appointments with potential customers based on a pre-prepared list or database. "Elevator pitch talks for making an appointment" are concise and persuasive sales talks that capture the attention and interest of potential customers within a short time frame (for example, while riding in an elevator), and they focus on conveying the main benefits and value of a product or service in a condensed form.

[0035] "Automatic appointment-making conversation script generation" is the process of generating a conversation script for making appointments based on specific customer data and situations. It creates a personalized conversation tailored to each customer's needs and interests, allowing sales representatives to use it in actual conversations, thereby supporting efficient appointment-making. Specifically, "Automatic appointment-making conversation script generation" includes basic appointment-making conversation script generation, standard elevator pitch conversation generation, automatic solution success case extraction, automatic solution-appeal conversation generation (features, benefits, value, etc. are shown here), automatic solution-appeal conversation generation for handling objections, elevator pitch conversation for hidden needs when past needs exist, and AI-generated elevator pitch conversation for hidden needs when past needs do not exist.

[0036] The role-playing model also includes interview talks, which include hero (heroine) make-up talk, opening value talk, collection of needs behind the needs, solution needs hearing talk, elevator pitch, elevator pitch of the needs behind the needs, value template collection, encouraging objection handling collection, solution (success story) collection, solution reverse calculation type SPIN, and solution reverse calculation type proposal presentation.

[0037] Furthermore, the role-playing model generation unit 114 preferably includes an automatic solution matching function and an automatic role-playing suggestion function, which are functions closely related to the role-playing model. The "automatic solution matching function" analyzes customer needs and background information and automatically recommends optimal products and services (solutions). This enables sales representatives to propose the most appropriate solution for each customer. The "automatic role-playing suggestion function" automatically generates talk scenarios and scripts based on the recommended solutions to mimic the flow of actual sales conversations. Specifically, the automatic role-playing suggestion function automatically generates a role-playing model by inputting a single customer's name, generating a hero (or heroine) make-up talk, an opening value talk, a collection of needs behind the needs, a solution needs hearing talk, an elevator pitch, an elevator pitch for the needs behind the needs, a value template collection, a collection of encouraging counterarguments, a collection of solutions (success stories), a solution reverse-calculation SPIN, and a solution reverse-calculation proposal presentation.

[0038] A "large-scale language model" is a type of probabilistic model used in natural language processing, which is a model for probabilistically predicting how likely a given word or sentence is to occur in natural language. Specifically, a language model calculates the occurrence probability of a given word sequence or sentence, or compares the occurrence probabilities of multiple word sequences or sentences, making it possible to automatically generate the most likely word or sentence based on the context when predicting the next word or sentence, or to generate a sentence that meets specific conditions.

[0039] Furthermore, the large-scale language model is specialized for creating assumed targets and role-playing models. As a result, the large-scale language model unit 113 can improve generation accuracy and efficiency by specializing in creating assumed targets and role-playing models using natural language processing based on target information in a marketing database. Specifically, by focusing on learning vocabulary and grammar related to target information, more realistic and specific assumed targets and role-playing models can be generated, and by optimizing the generation process, they can be generated in a shorter time. Furthermore, the large-scale language model unit 113 generates assumed targets and role-playing models based on new target information, enabling customization to specific targets, thereby enabling more effective sales activities. Specifically, assumed targets and role-playing models can be generated based on target elements such as age, gender, occupation, interests, lifestyle, values, needs, and challenges, and can include specialized terminology and knowledge tailored to specific companies, products, and services.

[0040] As described above, the marketing activity support device 1 includes a marketing database 121 in which target information consisting of a plurality of target elements related to targets in marketing activities, talk scripts for the targets, and success / failure information indicating the success or failure of marketing activities according to the talk scripts are stored in association with each other, an information receiving unit 111 that receives new target information consisting of one or more target elements, a talk script extraction unit 112 that extracts talk scripts that have made marketing activities successful based on the new target information and the target information and success / failure information in the marketing database 121, and a natural language processing unit 113 that performs one or more types of natural language processing based on the target information in the marketing database 121. By combining the above, it is possible to probabilistically predict how likely the words and sentences given in the prompt are to occur in natural language, and learning and adjustment are performed to generate a hypothetical target and role-playing model through sentence generation and question answering, resulting in a configuration comprising a large-scale language model unit 113 having a large-scale language model specialized for creating hypothetical targets and role-playing models, and a role-playing model generation unit 114 which generates a hypothetical target based on new target information by outputting a prompt to the large-scale language model unit 113, and generates a role-playing model when the talk script extracted by the talk script extraction unit 112 is applied to the hypothetical target.

[0041] According to the above configuration, as shown in FIG. 2, in the marketing activity support device 1, the large-scale language model unit 113 learns using information including each target element in the marketing database 121 having each database 1211 to 1216, and the new target information in the information receiving unit 111 and the talk script in the talk script extraction unit 112 correspond to explanatory variables in machine learning, and in response to a prompt from the role-playing model generation unit 114, the large-scale language model unit 113 generates a role-playing model as equivalent to the target variable in machine learning.

[0042] As a result, the marketing activity support device 1 receives new target information, extracts an optimal talk script from past successes, and generates a realistic role-playing model based on the talk script using a large-scale language model specialized for creating hypothetical targets and role-playing models, thereby enabling sufficient support for marketing activities. In particular, by using a large-scale language model to generate role-playing models rather than ordinary machine learning, it is possible to generate new content and variations not included in the training data, making it possible to generate flexible and diverse role-playing models tailored to customers and situations, as well as to generate creative content not included in the training data.

[0043] Furthermore, because large-scale language models can improve their accuracy through training, they can learn from past successes and continuously generate more effective role-playing models. In other words, they can analyze the results of sales activities and generate role-playing models that reflect areas for improvement, or they can reflect customer feedback and generate role-playing models that increase customer satisfaction.

[0044] In this way, by using a large-scale language model, the marketing activity support device 1 provides advanced functions that are difficult to achieve with conventional machine learning approaches. Specifically, it utilizes the unique capabilities of large-scale language models, such as complex text generation using natural language processing and context-based content generation. Conventional machine learning models generally perform prediction, classification, clustering, etc., mainly using numerical data and categorical data, and are limited in terms of processing large-scale text data and generating new text. On the other hand, large-scale language models are good at learning language patterns from large amounts of text data and generating new text based on given prompts. Therefore, in situations that require the use of diverse and complex language, such as role-playing models, an approach using a large-scale language model is more appropriate, and it is possible to provide functions that are difficult to achieve with conventional machine learning methods alone.

[0045] As shown in FIG. 3, the marketing activity support device 1 includes an information receiving unit 111, a talk script extraction unit 112, a large-scale language model unit 113, and a role-playing model generation unit 114, as well as a talk script generation unit 115, a target information acquisition unit 116, a database update unit 117, a terminal control unit 118, and a communication unit 13. Each unit 111 to 118, except for the communication unit 13, is included in the control unit 11, which is a computer. Some or all of the units included in the control unit 11 may be configured as either hardware or software. Input to the information receiving unit 111 is performed via an input unit 22, such as a keyboard, of a user terminal 2 operated by a user 4. Furthermore, the role-playing model created by the large-scale language model unit 113 and the role-playing model generation unit 114 is displayed on an output unit 21, such as a display, of the user terminal 2.

[0046] The communication unit 13 is connected to a user terminal 2 operated by a user 4 via the Internet 5 so as to enable data communication. Data communication between the communication unit 13 and the user terminal 2 is not limited to the Internet 5, but may be via an information communication network such as a dedicated line or a local area network (LAN), or via Bluetooth (registered trademark), a wireless communication standard for short-range communication. Data communication between the communication unit 13 and the user terminal 2 may be via a dedicated line to prevent information leakage. Furthermore, when communicating data via the Internet 5, the communication unit 13 and the user terminal 2 preferably have a virtual private network (VPN) function. If the communication unit 13 and the user terminal 2 incorporate a VPN function, data communication from the user terminal 2 passes through a VPN connection, thereby protecting the data from unauthorized external access and ensuring secure communication. The user terminal 2 and the administrator terminal 7 are information processing devices such as general personal computers, laptop computers, smartphones, and tablet terminals.

[0047] The terminal control unit 118 has a function of setting the display screen of the user terminal 2 to a screen suitable for creating a role-playing model. For example, as shown in Fig. 4, the terminal control unit 118 forms a role-playing model creation screen 211 for inputting information necessary for creating a role-playing model on the display screen that is the output unit 21 of the user terminal 2. The role-playing model creation screen 211 has an area for inputting information about the customer who will perform role-playing as target information, and allows input of information such as name, age, sex, occupation, position, telephone number, and email address.

[0048] The display screen of the user terminal 2 also includes a question box 212 and an answer box 213, which function as an interface for interaction with the large-scale language model unit 113. The question box 212 receives prompts such as questions or instructions from the user 4 to the large-scale language model unit 113. The answer box 213 displays appropriate answers or generated content based on the input from the question box 212. For example, if the user 4 inputs into the question box 212, "I don't know what exactly to write and how to write it for the performance indicators. I'd appreciate some advice," the answer box 213 outputs, "First, let's clarify the purpose of setting the performance indicators. For example, possible purposes include improving the efficiency of sales activities, improving customer satisfaction, and increasing sales. Once the purpose is clear, it will be easier to select appropriate indicators." This allows the terminal control unit 118 to provide the user 4 with various input examples for marketing activities through interactions between the user 4 and the large-scale language model unit 113.

[0049] As shown in FIG. 3 , the marketing activity support apparatus 1 includes an input device 15 connected to an input receiving unit 1192 and a display device 14 connected to a display control unit 1191. Examples of the input device 15 include a keyboard, a mouse, a touch panel, and a voice input device. Examples of the display device 14 include a liquid crystal display device. This allows the marketing activity support apparatus 1 to be configured using an information processing device such as a general personal computer, a laptop computer, a smartphone, or a tablet terminal. The marketing activity support apparatus 1 may lack at least one of the input device 15 and the display device 14. In this case, the input device 15 and the display device 14 are provided in an external terminal (not shown), allowing the marketing activity support apparatus 1 to be used as a role-playing model creation server. Furthermore, the marketing activity support apparatus 1 may be used as a user terminal 2, and a role-playing model may be created in the marketing activity support apparatus 1 while the user 4 receives support information generated by the large-scale language model unit 113.

[0050] In this embodiment, the case where the functions of the marketing activity support device 1 are installed in an information processing device is described, but the present invention is not limited to this, and the functions of the marketing activity support device 1 may be installed in cloud computing. In this case, cloud computing allows computer resources to be added as needed, making it possible to process large amounts of unique operation information and common operation information, enabling faster and more efficient processing, and also making it possible to easily expand processing capacity to accommodate a significant increase in the number of users 4.

[0051] Furthermore, the marketing activity support device 1 has a target information acquisition unit 116 that acquires target information accumulated when a sales entity conducts marketing activities, and a database update unit 117 that updates the marketing database 121 every time target information is acquired by the target information acquisition unit 116. This allows new target information obtained as a result of marketing activities to be reflected in the database in real time, making it possible to always formulate marketing strategies based on the latest information.

[0052] (Variation) As shown in Figure 5, the marketing activity support device 1 has a talk script generation unit 115 that generates a talk script that is likely to make marketing activities for new target information more successful based on the talk script extracted by the talk script extraction unit 112 and the new target information by outputting a prompt to the large-scale language model unit 113, and the role-playing model generation unit 114 may be configured to generate a role-playing model when the talk script generated by the talk script generation unit 115 is applied to a hypothetical target instead of the talk script extracted by the talk script extraction unit 112.

[0053] According to the above configuration, by outputting a prompt to the large-scale language model unit 113, the talk script generation unit 115 generates a talk script that is likely to make marketing activities for the new target information more successful, based on the talk script extracted by the talk script extraction unit 112 and the new target information. This allows the marketing activity support device 1 to generate a role-playing model that realizes more effective marketing activities.

[0054] In addition, the marketing activity support device 1 may have a talk script generation unit 115 that, when multiple talk scripts are extracted by the talk script extraction unit 112, outputs a prompt to the large-scale language model unit 113 to generate a talk script that is likely to make marketing activities for new target information successful based on the multiple talk scripts.

[0055] According to the above configuration, multiple existing talk scripts can be utilized to efficiently generate customized talk scripts that are optimal for new targets, thereby further enhancing the effectiveness of marketing activities.

[0056] (Marketing Database 121) Next, the marketing database 121 will be described in detail. The marketing database 121 is stored in the memory unit 12. The memory unit 12 is connected to the control unit 11 so that data can be communicated therewith. The memory unit 12 may be configured with a hard disk, or may be configured with a combination of a hard disk and memory. In the case of a configuration that combines a hard disk and memory, some of the data and indexes used by the database can be cached in the memory as needed, thereby speeding up access to the database. The memory unit 12 may be a data server that is connected to an information communication network such as the Internet 5, separate from the marketing activity support device 1. The memory unit 12 may also be configured with multiple data servers, one for each database.

[0057] As shown in FIGS. 6 and 7, the marketing database 121 includes a customer information database 1211, a solution database 1212, a sales talk database 1213, a role-playing database 1214, a results database 1215, and other databases 1216.

[0058] The customer information database 1211 has target element items such as a customer ID, name, company name, job title, telephone number, email address, address, products / services of interest, and purchase history, and each item is associated with a target element and stored to form one piece of target information. For example, when the customer ID is "1," the target information stored is: name: Taro Yamada, company name: A Co., Ltd., job title: Sales, telephone number: 090-1234-5678, email address: BBB, address: Minato-ku, Tokyo, products / services of interest: Product X, Service Y, and purchase history: "January 2023: Product X, March 2023: Service Y."

[0059] As a result, the customer information database 1211 aggregates and stores detailed information about each customer, enabling a personalized approach in marketing activities. Furthermore, by utilizing targeting elements such as customer IDs, personal information, contact information, and purchase history recorded in the customer information database 1211, it is possible to deeply understand the customer's needs and preferences and develop an effective communication strategy based on that understanding. For example, it is possible to understand what products and services a customer is interested in based on their past purchase history and use this information to propose similar or related new products. Furthermore, sales entities can select the most appropriate approach for each customer based on the customer's job title or company name, enabling the development of an efficient and targeted marketing strategy. In this way, the customer information database 1211 makes it possible to implement customized marketing tailored to each individual customer, aiming to improve customer satisfaction and maximize sales opportunities.

[0060] The solution database 1212 has target element items such as a solution ID, solution name, summary, target customer, value provided, price, and implementation examples, and stores target elements in association with each item to form one piece of solution information. For example, when the solution ID is "1", solution information is stored that includes a solution name of Product X, a summary of a customer management system, a target customer of small and medium-sized enterprises, value provided as sales efficiency, a price of 100,000 yen / month, and implementation examples of Co., Ltd. A and Co., Ltd. B.

[0061] This allows the solution database 1212 to centrally manage detailed information about each solution offered by a company. Specifically, by storing items such as the solution ID, solution name, overview, target customers, value provided, price, and case studies, sales and marketing personnel can quickly identify and propose optimal products and services to customers. For example, a solution that improves sales efficiency is remembered as providing value particularly to small and medium-sized enterprises (SMEs), making it effective when approaching customers in this segment. Furthermore, combining price information with case studies allows for concrete demonstration of the cost performance and actual benefits for customers, enabling a credible sales pitch. Including case studies allows potential customers to see examples of product success and specifically communicate their benefits, making it a powerful tool for favorably advancing sales negotiations. In this way, the solution database 1212 contributes to sales promotion and improved customer satisfaction by clarifying the features of products and services and supporting appropriate matching with customers.

[0062] The sales talk database 1213 has target element items such as a talk ID, situation, target, purpose, and talk content, and each item is associated with a target element and stored to form one piece of solution information. For example, when the talk ID is "1," sales talk information is stored in which the situation is a first visit, the target is a small business owner, the purpose is proposing the introduction of product X, and the talk content is "an overview of product X, an introduction case, and an explanation of the benefits of introducing it."

[0063] This makes it possible for the sales talk database 1213 to provide effective talk scripts tailored to sales scenarios. By storing target element items such as talk ID, situation, target purpose, and talk content, the sales entity can clearly understand what kind of conversation should be developed when approaching a specific customer. For example, in a conversation with the manager of a small or medium-sized business being visited for the first time, the sales entity can communicate efficiently without having to prepare content tailored to the scenario, as the sales talk database 1213 stores talk content proposing the introduction of product X and explaining the product's overview, introduction cases, and benefits obtained by introducing the product. This allows the sales entity to shorten preparation time and realize personalized dialogue tailored to the customer.

[0064] The role-playing database 1214 has target element items such as role-playing ID, situation, target, purpose, and role-playing content, and each item is associated with a target element and stored to form one piece of sales talk information. For example, if the role-playing ID is "1," sales talk information is stored in which the situation is a first visit, the target is a small business owner, the purpose is proposing the introduction of product X, and the role-playing content is "role-playing in which the sales representative and customer roles are played and the customer role is proposed to introduce product X."

[0065] The role-playing database 1214 thus enables salespeople to practice various sales scenarios in advance, thereby improving their performance during actual customer interactions. The role-playing database 1214 provides salespeople with opportunities to learn the flow of dialogue and develop feedback-handling skills through role-playing based on actual customer-facing scenarios. Information including the role-playing ID, situation, target, purpose, and role-playing content helps salespeople hone detailed communication skills, such as the choice of words and timing of responses used in actual conversations with customers. For example, in a situation where a salesperson is proposing product X during an initial meeting with a small-business owner, the salesperson can refer to an appropriate role-playing script from this database and simulate how to handle various questions and objections in advance. This allows the salesperson to approach actual sales situations with greater confidence, which in turn increases the likelihood of gaining customer trust and promoting the product.

[0066] The results database 1215 has target element items such as a result ID, solution ID, customer ID, talk ID, role-playing ID, implementation time, result indicator, and result content, and each item is associated with a target element and stored to form one sales talk information. For example, if the result ID is "1", the following result information is stored: solution ID 1, customer ID 1, talk ID 1, role-playing ID 1, implementation time January 2023, result indicator sales, and result content 10% increase in sales.

[0067] As a result, the results database 1215 makes it possible to track and analyze sales activities and their results based on specific figures and indicators. The results database 1215 records the results at each step from proposing a solution to closing a deal, making it possible to clarify which parts of the sales process are successful or need improvement. By including data such as result ID, solution ID, customer ID, talk ID, role-playing ID, implementation time, result indicators, and result content, the effectiveness of sales strategies can be measured and the extent to which specific solutions or sales approaches are accepted by customers can be understood in concrete figures. For example, in cases where results such as increased sales are achieved, it is possible to analyze what brought about that result, which becomes a valuable source of information for replicating successful sales processes or applying them to other cases.

[0068] Other databases 1216 store industry data, job data, market data, competitive data, etc. Industry data includes information on the trends, size, growth rate, major players, and trends within a specific industry, enabling sales entities to understand industry-specific needs and propose products and services tailored to that industry. Job data provides information related to specific job types and job descriptions, helping to gain deeper insight into target customers' daily tasks and the challenges they face. Market data includes market size, growth opportunities, customer behavior patterns, demand trends, etc., and is an important indicator in formulating marketing strategies. Competitive data provides information on competitors' products and services, pricing, market share, strengths and weaknesses, marketing strategies, etc., helping to determine a company's positioning and develop strategies to gain competitive advantage.

[0069] (Marketing activity support program) 3, in the marketing activity support device 1, the information receiving unit 111, talk script extraction unit 112, large-scale language model unit 113, role-playing model generation unit 114, talk script generation unit 115, target information acquisition unit 116, database update unit 117, and terminal control unit 118 may be configured as either hardware or software. Each of these units 111 to 118 constitutes at least a part of the control unit 11. When each of the units 111 to 118 is configured as software, the control unit 11 is configured to execute a marketing activity support program that generates a role-playing model by utilizing a language model for natural language processing specialized for creating a role-playing model.

[0070] To give a specific example using FIG. 8, the marketing activity support program includes an information receiving process step (S1) for receiving new target information consisting of one or more target elements in a computer (control unit 11) of the marketing activity support device 1 having a marketing database 121 in which target information consisting of multiple target elements related to targets in marketing activities, talk scripts for the targets, and success / failure information indicating the success or failure of marketing activities based on the talk scripts are stored in association with each other; a talk script extraction process step (S2) for extracting talk scripts that have made marketing activities successful based on the new target information, the target information in the marketing database 121, and the success / failure information in the marketing database 121; The program executes a large-scale language model processing step (S3) having a large-scale language model specialized for creating an assumed target and a role-playing model by combining one or more types of natural language processing to probabilistically predict how likely words and sentences given in a prompt are to occur in natural language, and learning and adjustment are performed to generate an assumed target and a role-playing model through sentence generation and question answering, and a role-playing model generation processing step (S4) that generates an assumed target based on new target information by outputting a prompt to the large-scale language model processing step (S3) and generates a role-playing model when the talk script extracted in the talk script extraction processing step (S2) is applied to the assumed target.The marketing activity support program may be configured to cause a computer to execute the functions of each of the units 111 to 118 in the control unit 11 as processing steps.

[0071] According to the above configuration, it is possible to fully support marketing activities by receiving new target information, extracting an optimal talk script from past successes, and using a large-scale language model specialized in creating hypothetical targets and role-playing models to generate realistic role-playing models based on the talk script. In particular, by using a large-scale language model to generate role-playing models rather than ordinary machine learning, it is possible to generate new content and variations not included in the training data, making it possible to flexibly generate diverse role-playing models tailored to customers and situations, and to generate creative content not included in the training data.

[0072] Furthermore, simply by installing the marketing activity support program in an information processing device such as a personal computer or tablet terminal, the information processing device can function as the marketing activity support device 1. The program may be distributed in a state recorded on a computer-readable recording medium such as a CD-ROM or USB memory, or may be distributed via a two-way communication network or communication line such as the Internet or a one-way communication network such as a television broadcast.

[0073] (Marketing activity support method) The marketing activity support device 1 is configured to cause a computer (control unit 11) to execute a marketing activity support method. Specifically, the computer (control unit 11) of the marketing activity support device 1 has a marketing database 121 in which target information consisting of a plurality of target elements related to targets in marketing activities, talk scripts for the targets, and success / failure information indicating the success or failure of marketing activities according to the talk scripts are stored in association with each other, and the computer (control unit 11) of the marketing activity support device 1 is configured to execute the following processes: an information receiving process for receiving new target information consisting of one or more target elements; a talk script extraction process for extracting talk scripts that have made marketing activities successful based on the new target information and the target information and success / failure information in the marketing database 121; This method combines one or more types of language processing to probabilistically predict how likely words and sentences given in prompts are to occur in natural language, and performs learning and adjustment to generate an assumed target and role-playing model through sentence generation and question answering, thereby executing large-scale language model processing having a large-scale language model specialized for creating an assumed target and role-playing model, and role-playing model generation processing that generates an assumed target based on new target information by outputting a prompt to the large-scale language model processing, and generates a role-playing model when the talk script extracted in the talk script extraction processing is applied to the assumed target. Note that the marketing activity support method may be configured to have a computer execute the functions of each of the units 111 to 118 in the control unit 11 as processing steps.

[0074] It should be noted that within the scope of the concept of the present invention, those skilled in the art may conceive of various modifications and alterations. Therefore, it is understood that such modifications and alterations fall within the scope of the present invention. For example, even if a person skilled in the art appropriately adds, deletes, or modifies components of the above-described embodiment, or adds, omits, or modifies the conditions of a process, such modifications are also included within the scope of the present invention as long as they maintain the gist of the present invention. [Explanation of symbols]

[0075] 1 Marketing activity support device 2. User terminal 4 User 5. Internet 7 Administrator terminal 11 Control section 12 Storage section 13 Communications Department 14 Display device 15 Input Devices 111 Information Reception Department 112 Talk script extraction unit 113 Large-scale Language Modeling 114 Role-playing model generation unit 115 Talk script generation unit 116 Target Information Acquisition Unit 117 Database Update Department 118 Terminal control unit 121 Marketing Database

Claims

1. a marketing database in which target information consisting of a plurality of target elements relating to a target in a marketing activity, a talk script for the target, and success / failure information indicating the success or failure of the marketing activity according to the talk script are stored in association with each other; an information receiving unit that receives new target information consisting of one or more target elements; a talk script extraction unit that extracts the talk script that made the marketing activity successful based on the new target information, the target information in the marketing database, and the success / failure information; a large-scale language model unit that combines one or more types of natural language processing based on the target information in the marketing database to probabilistically predict how likely words and sentences given in the prompt are to occur in natural language, and that performs learning and adjustment to generate a hypothetical target and a role-playing model through sentence generation and question answering; and a role-playing model generation unit that generates the assumed target based on the new target information by outputting the prompt to the large-scale language model unit, and generates the role-playing model when the talk script extracted by the talk script extraction unit is applied to the assumed target; A marketing activity support device having the above.

2. The talk script extraction unit Using the target information, the talk script, and the success / failure information as training data, machine learning is performed using a talk script prediction model trained to predict the successful talk script based on the target information and the success / failure information, and a talk script that is likely to make the marketing activity successful for the new target information is extracted. The marketing activity support device according to claim 1 .

3. a talk script generation unit that generates a talk script that is likely to make the marketing activity for the new target information more successful, based on the talk script extracted by the talk script extraction unit and the new target information, by outputting the prompt to the large-scale language model unit; The role-playing model generation unit generates the role-playing model when the talk script generated by the talk script generation unit is applied to the assumed target instead of the talk script extracted by the talk script extraction unit. The marketing activity support device according to claim 2 .

4. a talk script generation unit that generates a talk script that is likely to make the marketing activity for the new target information successful based on the plurality of talk scripts by outputting the prompt to the large-scale language model unit when the plurality of talk scripts are extracted by the talk script extraction unit; The marketing activity support device according to claim 1 .

5. a target information acquisition unit that acquires the target information accumulated when a sales entity carries out the marketing activity; a database update unit that updates the marketing database every time the target information is acquired by the target information acquisition unit; 2. The marketing activity support device according to claim 1, further comprising:

Images