Generation apparatus, generation method, and generation program

Abstract

This enables users to consider appropriate purchasing decisions. [Solution] The generation device 100 receives prompts expressed in natural language text, which include information about the product selected by a user who wants to purchase the product, information about the user's situation, and a command to generate the target person's opinion on the product, into a large-scale language model that has information about the target person's profile set up, and generates the target person's opinion on the product. The generation device 100 outputs the generated opinion of the target person on the product.

Description

Technical Field

[0001] The present invention relates to a generation device, a generation method, and a generation program.

Background Art

[0002] A user who is considering purchasing a product or service can access a website constructed by the Internet via a terminal device or the like operated by the user, and can conduct online purchase consideration or actual purchase.

[0003] When a user purchases a product in an online format, the user may refer to the product description. However, simply presenting a product description or the like makes it difficult to arouse the user's purchasing desire. Therefore, in order to clearly explain the advantages of a product to a user, a conventional technique of displaying an avatar that performs an explanatory operation of a target product or a photographed image object of a real person is known (for example, see Patent Document 1).

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, in the prior art, it may be difficult for a user to appropriately consider a purchase. For example, the prior art is a technique for explaining a product using an avatar or an image of a real person. However, the prior art only introduces a pre-registered product description to the user, and does not consider the user's situation or the relationship with other persons assumed by the user when purchasing a product, and there are problems in appropriately and efficiently selecting a purchased product.

Means for Solving the Problems

[0006] Therefore, in order to solve the above-mentioned problems and achieve the objective, the present invention is characterized by comprising: a generation unit that receives prompts expressed in natural language text, which include information about a product selected by a user who wishes to purchase the product, information about the user who wishes to purchase the product, and a command to generate the opinion of the target person regarding the product, to a large-scale language model in which information about the profile of the target person is set, and generates the opinion of the target person regarding the product; and an output unit that outputs the opinion of the target person regarding the product generated by the generation unit. [Effects of the Invention]

[0007] The present invention has the effect of enabling users to consider appropriate purchasing decisions. [Brief explanation of the drawing]

[0008] [Figure 1] Figure 1 is a diagram illustrating the overall process of the generating apparatus according to the embodiment. [Figure 2] Figure 2 shows the configuration of the generating apparatus according to this embodiment. [Figure 3] Figure 3 is a table diagram showing an example of user information according to the embodiment. [Figure 4] Figure 4 is a table diagram showing an example of person information according to the embodiment. [Figure 5] Figure 5 is a table diagram showing an example of product-related information according to the embodiment. [Figure 6] Figure 6 shows an example of a prompt according to the present invention. [Figure 7] Figure 7 shows an example of the opinion aggregation process according to the embodiment. [Figure 8] Figure 8 shows an example of the generation process according to the embodiment. [Figure 9] Figure 9 shows an example of a display screen according to the embodiment. [Figure 10]Figure 10 is a flowchart showing the processing performed by the generation apparatus according to the embodiment. [Figure 11] Figure 11 shows an example of a computer that realizes the generation apparatus according to the embodiment. [Modes for carrying out the invention]

[0009] The embodiments for carrying out the present invention (hereinafter referred to as "embodiments") will be described below with reference to the drawings. However, the embodiments are not limited to those described below.

[0010] <Overview> (background) A technology is known that enables users to purchase goods and services (hereinafter sometimes simply referred to as "goods") online by virtually displaying them on a website built on the internet. Furthermore, a reference technology is known that displays an avatar or a photographic object of a real person performing explanatory actions about the product in order to stimulate the user's desire to purchase when they are considering purchasing a product.

[0011] However, the above-mentioned reference technology only presents pre-registered product descriptions to the user using avatars or images of real people. Therefore, the reference technology does not take into account the circumstances under which the user selects a product, the person to whom the user is giving the product as a gift, or the relationship between the user and the recipient (the user's situation), and thus has challenges in selecting products appropriately and efficiently.

[0012] (Processing by generation device 100) Therefore, when a user purchases a product, the generation device 100 according to this embodiment generates the target person's opinion on the product based on a large-scale language model in which the target person's profile has been set using information about the target person's profile, and outputs this opinion to the user using information about the product and information about the user's situation who desires to purchase the product.

[0013] In addition, in the present embodiment, an example in which the generation device 100 generates the opinion of the target person will be described in a situation where the user selects and purchases a product for giving a present or the like to the target person (for example, the user's mother, father, sibling, friend, colleague or supervisor at work, etc.). However, since the situation in the above-described present embodiment is merely an example, other situations are of course included.

[0014] In addition, the "information on the person image" is persona information such as the gender, age, personality, values, etc. of the target person, and may hereinafter be simply referred to as "person information" in some cases. The "information on the product" is information including product descriptions and detailed information on the product selected by the user who desires to purchase the product, and may hereinafter be simply referred to as "product-related information" in some cases. Further, the "information on the situation of the user who desires to purchase the product" is information representing the relationship between the user and the target person to whom a present or the like is given when the user purchases the product, the situation of product selection, etc., and may hereinafter be simply referred to as "user situation information" in some cases.

[0015] Here, the overall image of the processing by the generation device 100 will be described. FIG. 1 is a diagram for explaining the overall image of the processing of the generation device 100 according to the embodiment. The generation device 100 shown in FIG. 1 is an example of a computer that provides a technique for realizing the information processing described below.

[0016] The generation device 100 uses the person information of the target person ((1-1) in FIG. 1) to set the person image such as the gender, age, personality, values, etc. of the person for the large language model ((1-2) in FIG. 1). Here, the above-mentioned "setting of the person image" means setting (pre-training) the person image (persona) of the target person for the large language model so that the opinion on the product generated by the large language model becomes a virtual opinion based on the person image of the target person.

[0017] The generation device 100 receives prompts expressed in natural language text, which include product-related information for a product selected by a user who wishes to purchase a product, user situation information, and a command to generate the target person's opinion on the product, into a large-scale language model in which the profile of the target person has been set, and generates the target person's opinion on the product (Figure 1 (2-1) to (2-3)).

[0018] The generation device 100 then outputs the subject person's opinion on the generated product (Figure 1 (3-1)). For example, the generation device 100 can output the subject person's opinion as a numerical score representing satisfaction with the product, the subject person's impressions of the product, or an opinion that aggregates the opinions of multiple people (aggregated opinion) (Figure 1 (3-2)).

[0019] In this way, the generation device 100 according to this embodiment can provide feedback to the user regarding opinions and evaluations of products based on the profile of the target person when the user is selecting or purchasing a product. In other words, the generation device 100 makes it possible for the user to know in advance whether the target person will like the product when the user is purchasing it. As a result, the generation device 100 has the effect of facilitating appropriate purchasing considerations by the user.

[0020] <Description of Generator 100> The configuration of the generation device 100 according to this embodiment will now be described. Figure 2 is a diagram showing the configuration of the generation device 100 according to this embodiment. As shown in Figure 2, the generation device 100 has a communication unit 110, a storage unit 120, and a control unit 130.

[0021] Although not shown in Figure 2, the generation device 100 may be equipped with an input unit such as a keyboard or mouse to receive input such as operations from an administrator. Furthermore, the generation device 100 may be equipped with a display or the like to show information for managing the operation of the generation device 100 to an administrator.

[0022] (Communications Department 110) The communication unit 110 performs data communication related to the input of user information, personal information of the target person, product-related information concerning products selected by the user, and user status information. The communication unit 110 also performs data communication related to the output of generated opinions of the target person and information concerning selected products. The communication unit 110 is implemented using a NIC (Network Interface Card), etc., and controls communication via telecommunication lines such as a LAN (Local Area Network) or the Internet. The communication unit 110 is connected to the network by wired or wireless connection as needed, and can send and receive information bidirectionally with terminal devices operated by the user.

[0023] (Storage unit 120) The memory unit 120 stores data and programs used for various processes by the control unit 130, as well as various data acquired as a result of the operation of the control unit 130. In addition, the memory unit 120 stores opinions of real people, etc., received by the reception unit 133, which will be described later.

[0024] The memory unit 120 is implemented using semiconductor memory elements such as RAM (Random Access Memory) and flash memory, or storage devices such as hard disks and optical discs. As shown in Figure 2, the memory unit 120 also includes a user information DB 121, a person information DB 122, a product-related information DB 123, and a generation model DB 124.

[0025] (User Information DB121) User Information DB121 is a database that stores information about users who wish to purchase products (user information). This user information is used by the generation unit 134 (described later) to identify a model in which the profile of the person targeted by the user selecting and purchasing the product is defined when the opinion of the target person is generated.

[0026] Specifically, the user information DB121 stores information such as information to identify the user (user identification information), information about the user's attributes (user attributes), information about the relationship between the user and the person in question (relationships), and information about the reason for purchasing the product (reason for purchase).

[0027] Here, an example of user information stored in the user information DB121 will be explained using Figure 3. Figure 3 is a table diagram showing an example of user information according to the embodiment. The user information DB121 stores information related to each item, such as "No," which is information that identifies individual data included in the user information, "user identification information," "user attributes," "relationships," and "reason for purchase," in a table format or the like.

[0028] For example, as shown in Figure 3, the user information DB121 can store user information such as user identification information "User A" identified by No. "1", user attribute "Attribute a", relationship "User B (mother), User C (father), User D (younger brother)", and purchase reason "Purchased a gift for family".

[0029] The "user identification information" mentioned above refers to information that identifies a user who wishes to purchase a product, and includes, for example, information expressed by a combination of predetermined strings, numbers, symbols, etc. Note that the user identification information may be modified by deleting or replacing any information that could identify the individual user using publicly known technologies.

[0030] "User attributes" refer to attribute information of users who wish to purchase a product, and include information such as age, gender, place of residence, annual income, occupation, educational background, hobbies and preferences, and purchase history. "Relationships" refer to information that represents the relationship between the user and the person in question, and include information that shows relationships between family members such as mother, father, and siblings, as well as information that shows connections with the user in friendships and workplace relationships. "Reasons for purchase" refer to information about the reasons why the user wishes to purchase a product, and include information about reasons for purchasing the product such as "birthday present" or "gift for employment / retirement."

[0031] (Person information DB122) Person Information DB122 is a database that stores information about the subject person, such as their gender, age, personality, and values ​​(person information). Specifically, Person Information DB122 stores information for identifying the subject person (person identification information), information indicating which user the person is associated with (corresponding user), information indicating gender, age, and whether or not it is the person in question (person attributes), information about the person's preferences as answered by the person (survey information), and information about the person's past purchase history (purchase information).

[0032] Here, an example of person information stored in the person information DB122 will be explained using Figure 4. Figure 4 is a table diagram showing an example of person information according to the embodiment. The person information DB122 stores information related to each of the following items in a table format, such as "No," which is information that identifies individual data included in the person information, "person identification information," "corresponding user," "person attributes," "survey information," and "purchase information."

[0033] For example, person information DB122 can store person identification information "User B (mother)" identified by No. "1", corresponding user "User A", person attributes "Gender and age: Female in her 60s, Is it the person herself or not: Yes", survey information "Hobbies and preferences: Gardening, cooking, Taste: Does not like sweets, likes Japanese food, especially fish dishes", and purchase information "Purchase information b".

[0034] "Personal identification information" refers to information that identifies a person, and includes, for example, information expressed by a combination of predetermined strings of characters, numbers, symbols, etc. Note that personal identification information may be modified by deleting or replacing information that could identify the individual using publicly known technologies.

[0035] "Corresponding user" refers to information that represents the relationship between a user and a person, and includes information that represents the user corresponding to that person. For example, if the username etc. included in the "User Identification Information" shown in Figure 3 matches the username etc. included in the "Corresponding User" shown in Figure 4, it can be said that the user who makes a purchase etc. is associated with the person included in the person information.

[0036] "Personal attributes" refer to attribute information of the subject person or a person similar to the subject person, and include, for example, information such as age, gender, whether the registered information is about the person in question, place of residence, annual income, occupation, and educational background.

[0037] "Survey information" refers to the results of responses to a questionnaire conducted in advance to understand a person's preferences, etc., and includes information such as the person's hobbies, tastes, and preferences (likes / dislikes). "Purchase information" refers to historical information about purchases made by a designated person in the past, and includes information such as the name of the purchased product, the date of purchase, and the purchase price. The above survey information may include general survey results that do not contain personally identifiable information. Furthermore, there are no particular limitations on the method of conducting and compiling the survey; any method that can understand a person's preferences, etc., is acceptable for conducting and compiling the survey.

[0038] (Product-related information DB123) Product-related information DB123 is a database that stores information about products selected by users who wish to purchase products (product-related information). Specifically, product-related information DB123 stores information such as the product name (product name), product description (product description), price (price), and product image (image).

[0039] Here, an example of product-related information stored in the product-related information DB123 will be explained using Figure 5. Figure 5 is a table diagram showing an example of product-related information according to the embodiment. The product-related information DB123 stores information related to each item, such as "No," which is information that identifies individual data included in the product-related information, "product name," "product description," "price," and "image," in a table format or the like.

[0040] For example, as shown in Figure 5, the product-related information DB123 stores the product name "Chips Satsuma Vanilla," identified by No. "1," the product description "A snack in which the subtly sweet flavor of sweet potato and vanilla are perfectly matched. The creamy and fluffy texture, combined with the crispness of the chips, allows you to enjoy a new kind of dessert-like snack," the price "100 yen," and the image "Image E."

[0041] "Product" refers to information such as the product name that identifies a product displayed on a website, etc., and includes product names set by the product's seller, etc., such as "Chips Satsuma Vanilla." "Product description" refers to information that explains what kind of product a product is, such as the product's contents, features, taste, aroma, appearance, and recommended points.

[0042] "Price" refers to information regarding the selling price of a product displayed on a website, etc., and includes information such as a numerical value of the selling price based on a specified currency, such as "100 yen." "Image" refers to still image data or moving image data of a product displayed on a website, etc., and includes information such as the still image data or moving image data itself, or the directory name or URL (Uniform Resource Locator) used to identify the location where the image data is stored.

[0043] (Generative model DB124) The Generative Model DB124 is a database that stores predetermined generative models used to generate opinions from a target individual. For example, the Generative Model DB124 can store large-scale language models as generative models.

[0044] Specifically, the generation device 100 according to this embodiment can use at least one of the following as a large-scale language model: "ChatGPT®", a large-scale language model possessing general-purpose knowledge, and "tsuzumi®", a predetermined large-scale language model on which adapter tuning is performed (see, for example, References 1 and 2).

[0045] (Reference 1):ChatGPT(OpenAI),<URL:https: / / openai.com / chatgpt> ,<Searched on August 26, 2020> (Reference 2): NTT version of large-scale language model "tsuzumi",<URL:https: / / www.rd.ntt / research / LLM_tsuzumi.html> ,<Searched on August 26, 2020>

[0046] (Control unit 130) Now, let's return to Figure 2 and continue the explanation. The control unit 130 has an internal memory for temporarily storing programs and processing data that define various processing procedures of the generation device 100, and is realized by electronic circuits such as a CPU (Central Processing Unit) and an MPU (Micro Processing Unit), and integrated circuits such as an ASIC (Application Specific Integrated Circuit) and an FPGA (Field Programmable Gate Array). As shown in Figure 2, the control unit 130 has an acquisition unit 131, a learning unit 132, a receiving unit 133, a generation unit 134, and an output unit 135.

[0047] (Acquisition part 131) The acquisition unit 131 acquires information from an external information processing device or the like that the generation device 100 uses to generate the opinion of the target person.

[0048] Specifically, the acquisition unit 131 acquires user information such as user identification information, user attributes, user relationships, and reasons for purchase from an external information processing device that manages user information. The acquisition unit 131 then stores the acquired user information in the user information DB 121.

[0049] Furthermore, the acquisition unit 131 acquires person identification information, user identification information (user identification information) of the person corresponding to the user, the attributes of the person, survey information, purchase information, etc., from an external information processing device that manages person information obtained by conducting pre-surveys, etc. The acquisition unit 131 then stores the acquired person information in the person information DB 122.

[0050] Furthermore, the acquisition unit 131 acquires product-related information such as product name, product description, price, and image of a product that a user wishes to purchase from an external information processing device that manages a website or other online product display site. The acquisition unit 131 then stores the acquired product-related information in the product-related information DB 123.

[0051] (Learning Section 132) The learning unit 132 performs a learning process on a predetermined large-scale language model that the generation unit 134, described later, uses to generate the opinions of the target person.

[0052] The learning unit 132 uses person information about the target person (the target person) identified using user information ("relationships") related to users who wish to purchase the product to set up a profile of the target person for the large-scale language model. Specifically, the learning unit 132 uses at least one of the following as person information for the target person to set up a profile of the target person for the large-scale language model: the target person's gender and age ("person attributes" shown in Figure 3), the relationship between the target person and the user ("corresponding user" shown in Figure 3), the results of the survey given by the target person ("survey information" shown in Figure 3), the target person's purchase information ("purchase information" shown in Figure 3), and attribute information of people similar to the target person's profile ("person attributes" shown in Figure 3).

[0053] For example, the learning unit 132 identifies the target person "User B (Mother)" using "User B (Mother)" included in the "Relationships" section of the user information shown in Figure 3. Next, the learning unit 132 acquires information about User B (Mother) included in the person information shown in Figure 4, such as gender and age "Female in her 60s," and survey information "Hobbies and preferences: Gardening, cooking; Taste: Dislikes sweets, likes Japanese food, especially fish dishes." Then, the learning unit 132 inputs the acquired person information about User B (Mother) and prompts expressed in natural language text, which instruct the model to output opinions by imitating the actions and words of a person according to their personality, hobbies, and preferences, into a large-scale language model. This allows the large-scale language model to be pre-trained to set up a profile of "User B (Mother)."

[0054] For example, the learning unit 132 inputs prompts such as, "You are User B (Mother), the mother of User A. Using User B (Mother)'s character information, express your opinion as if you were User B (Mother)," to the large-scale language model, thereby setting up the profile of User B (Mother).

[0055] Furthermore, the learning unit 132 learns a large-scale language model using the opinions of real people received by the reception unit 133, which will be described later. For example, regarding "Chips Satsuma Vanilla," the learning unit 132 can update the large-scale language model by inputting a prompt expressed in natural language text, which includes actual feedback from user B (mother) such as "It's sweet, but delicious," and a command to update the large-scale language model, which has a profile of user B (mother) set based on the actual feedback, so that it generates opinions identical or similar to those of the real user B (mother).

[0056] Furthermore, the learning unit 132 can set a profile of the target person for a large-scale language model using attribute information of people similar to the target person's profile as described above, which is from similar segments and statistically compiled from the attribute information of multiple people (e.g., gender, age, hobbies, preferences, etc.). For example, if there is no personal information for the user's actual younger brother, the learning unit 132 can set a profile of a "virtual younger brother" based on attribute information similar to that brother (e.g., attribute information belonging to the segment of men in their 20s). Also, if the user cannot directly ask their boss what they like when giving a gift, the learning unit 132 can set a profile of a "virtual boss" based on attribute information similar to that boss (e.g., attribute information belonging to the segment of men in their 40s).

[0057] (Reception desk 133) The reception unit 133 receives opinions from real people about the product. For example, the reception unit 133 displays a question to the user such as, "What was the opinion of actual user B (mother) about this product?" The reception unit 133 then receives the answer to the displayed question and stores the opinion of the real person in the storage unit 120.

[0058] (Generation unit 134) The generation unit 134 inputs a prompt, which includes product-related information for the product selected by the user, user status information, and a command to generate the target person's opinion on the product, into a large-scale language model in which the target person's profile is set, and generates the target person's opinion on the product.

[0059] Specifically, the generation unit 134 inputs a prompt, which includes product-related information of a user-selected product (at least one of the following: product name, price, product description, and product image) and a command to generate the opinion of the target person, to a large-scale language model whose profile has been set by the learning unit 132. The generation unit 134 then generates the opinion of the target person according to the relationship between the user and the target person and the purchasing situation. A specific example of the generation process by the generation unit 134 will be explained in the section related to Figure 8.

[0060] (Explanation of an example prompt) Here, an example of a prompt used by the generation unit 134 will be explained with reference to Figure 6. Figure 6 is a diagram showing an example of a prompt according to the embodiment. Figure 6 shows an example of a prompt input to a large-scale language model in which the profile of the target person is set, and the opinion of each target person that is generated based on the input of the prompt.

[0061] The large-scale language model shown in Figure 6 includes a person profile "model" based on the person information of User B (mother), who is the mother of the user shown in Figure 4. <1> The setting is "Mother (female in her 60s)". In addition, the large-scale language model shown in Figure 6 includes a character profile "Model" based on the personal information of User C (father), who is the user's father. <2> The above "model" is set as "Father (60s male)". <1> "Mother (woman in her 60s)" and "Model" <2> "Father (male, 60s)" is a profile of a person set up for a large-scale language model based on the actual personal information of the user's mother and father.

[0062] Furthermore, the large-scale language model shown in Figure 6 includes a "model" of a person based on person information similar to that of User D (brother), who is the user's younger brother. <3> The character is set as "a man in his 20s (a fictional younger brother)". <3> The "20s Male (Virtual Brother)" is a profile of a "virtual brother" created based on attribute information similar to the actual brother of the user, since no personal information existed for the user's real brother. In other words, even if no personal information exists for the target person, the generator 100 can set up a virtual profile for the large-scale language model using personal information of a person similar to the target person.

[0063] The generation unit 134 inputs prompts, including commands to generate opinions for each target person, such as "Please tell us what kind of reactions the three people would likely have to each of the three flavors (please also assign scores)," as shown in Figure 6 (1), to a large-scale language model in which the profiles of the target people have been set.

[0064] Then, as shown in Figure 6 (2), the generation unit 134 generates a score and impressions of the product for each individual, representing the opinions of the target individuals. The "impressions" mentioned above are the impressions of the product that the target individual is expected to have, and include, for example, the feelings when purchasing or acquiring the product, and the coherent thoughts that came to the person's mind. The "score" is the result of the target individual assigning a score to the product, such as their satisfaction level.

[0065] Specifically, the generation unit 134 is "model <1> The system generates scores and comments such as "Satsuma Vanilla: 2 points 'It's sweet, but not bad,' Flame Chili Pepper: 4 points 'I wonder if it's too spicy,' Seaweed Wasabi Cheese: 8 points 'It's Japanese style and I like it'" as the opinion of a "mother (woman in her 60s)" (Figure 6 (2-1)). In addition, the generation unit 134 generates "model <2> The system generates scores and comments such as "Satsuma Vanilla: 4 points 'Too sweet, I don't like it'", "Flame Chili Pepper: 9 points 'This looks delicious!'", and "Seaweed Wasabi Cheese: 7 points 'The spiciness of the wasabi is a nice accent'" as the opinion of a "father (male in his 60s)" (Figure 6 (2-2)). In addition, the generation unit 134 generates "model <3> The system generates scores and comments from a "20-something male (fictional younger brother)," such as "Satsuma Vanilla: 7 points 'Surprisingly good! The sweetness is junk-food-like,' Flame Chili Pepper: 8 points 'Exciting and fun,' Seaweed Wasabi Cheese: 4 points 'A bit too Japanese, maybe not so great'" (Figure 6 (2-3)).

[0066] (Explanation of an example of opinion aggregation process) Furthermore, the generation unit 134 can aggregate the opinions of multiple individuals that have been generated. Here, an example of opinion aggregation processing by the generation unit 134 will be explained using Figure 7. Figure 7 is a diagram showing an example of opinion aggregation processing according to the embodiment.

[0067] Figure 7 shows an example of a process for aggregating the scores and comments for each individual generated in the example process shown in Figure 6. Note that the generation of scores and comments for each individual shown in Figure 7 is the same as in Figure 6, so some explanation is omitted.

[0068] The generation unit 134 generates scores and opinions for each target person regarding the product, similar to the process shown in Figure 6 (Figure 7(1)). Next, the generation unit 134 combines the generated scores and opinions for multiple target people with prompts containing commands to aggregate the scores and opinions for multiple target people into a large-scale language model (model). <4> The opinion aggregation model is used as input (Figure 7(2-1)), and aggregated scores and comments are generated (Figure 7(2-2)).

[0069] For example, as shown in (2-2) of Figure 7, the generation unit 134 generates the results of summarizing the opinions of three people regarding the flavors of "Satsuma Vanilla," "Flame Chili Pepper," and "Seaweed Wasabi Cheese."

[0070] For example, regarding "Satsuma Vanilla," the generation unit 134 generates summarized opinions such as, "Overall rating: 4.3 points. Opinions were divided on the sweetness; my father and mother felt it was a little too sweet, but my younger brother seemed to enjoy the sweetness. Overall, it is characterized by its sweet taste, and the rating is average."

[0071] Furthermore, regarding "Flame Chili Pepper," generation unit 134 generates summarized opinions such as, "Overall rating: 7 points. My father and brother enjoyed the spiciness and gave it a high rating. On the other hand, my mother felt it was too spicy. Overall, it is quite popular with those who like spicy food, but it may be too much for those who cannot tolerate spiciness."

[0072] Furthermore, regarding "Nori Wasabi Cheese," generation unit 134 generates summarized opinions such as, "Overall rating: 6.3 points. My father and mother seem to like the Japanese accent of this flavor. However, my younger brother felt it was too Japanese and gave it a low rating. It's a flavor where the evaluation will be divided depending on whether you like Japanese flavors or not."

[0073] (Output section 135) Returning to Figure 2, the explanation continues. The output unit 135 outputs the subject person's opinion on the product generated by the generation unit 134. An example of the output processing by the output unit 135 will be explained in the section on "Examples of Processing" below.

[0074] (An example of processing) From here, an example of processing by the generation device 100 will be explained using Figure 8. Figure 8 is a diagram showing an example of the generation process according to the embodiment.

[0075] The generation device 100 (learning unit) uses the target person's information (Figure 8 (1-1)) to set a profile of the person targeted by the user who will purchase the product for the large-scale language model (Figure 8 (1-2)). In the example shown in Figure 8, the generation device 100 (learning unit) sets "mother (female in her 60s) (Figure 8 (1-3))", "father (male in his 60s) (Figure 8 (1-4))", and "male in his 20s (virtual younger brother) (Figure 8 (1-5))" as profiles of the user's family members for the large-scale language model.

[0076] The generation device 100 (acquisition unit) acquires product-related information (Figure 8 (2-2)) for the "product" selected by the user using the cursor 10. Next, the generation device 100 (generation unit) converts the acquired product-related information into natural language text (Figure 8 (2-3)).

[0077] Next, the generation device 100 (generation unit) identifies the target user based on the relationships "User B (mother), User C (father), User D (younger brother)" included in the user information stored in the user information DB 121. The generation device 100 (generation unit) also retrieves the purchase reason "purchasing a gift for family" included in the user information stored in the user information DB 121. Then, the generation device 100 (generation unit) inputs a prompt containing the product-related information converted into text and a command to generate an opinion corresponding to the identified person and the purchase reason to a large-scale language model in which the profiles of the user's family members are set (Figure 8 (2-4)), and generates the opinion of the target person (Figure 8 (2-5)).

[0078] For example, as shown in (2-6) of Figure 8, the generating device 100 (generating unit) generates opinions from a mother (woman in her 60s) such as, "It goes well with my specialty dish," and "Score: 8 points." Also, as shown in (2-7) of Figure 8, the generating device 100 (generating unit) generates opinions from a father (man in his 60s) such as, "I like this!" and "Score: 9 points." As shown in (2-8) of Figure 8, the generating device 100 (generating unit) generates opinions from a man in his 20s (fictional younger brother) such as, "I'd prefer something more substantial," and "Score: 2 points."

[0079] Furthermore, the generation device 100 (generation unit) can aggregate the opinions of each individual to generate the "aggregated opinions" explained in Figure 7 (Figure 8 (3)). The generation device 100 (output unit) then outputs the generated scores and comments, or the aggregated opinions (Figure 8 (4)).

[0080] Here, an example of the output of the target person's opinion by the generation device 100 will be explained using Figure 9. Figure 9 is a diagram showing an example of the display screen according to the embodiment. In Figure 9, the usage scenario is "The daughter (user) gives a gift to her family," the family composition is "Mother in her 60s, father in his 60s, and younger brother in his 20s," and the user's request is "I want to give a gift that everyone will be happy with." Note that since there is no personal information for the "younger brother in his 20s," a person profile (a fictional younger brother) is set using personal information similar to that of the younger brother.

[0081] For example, when a user selects "product d" using cursor 10, the generation device 100 (output unit) displays the opinions of each person, generated based on a large-scale language model in which the user's family members' profiles have been pre-configured, near the area where product d is displayed (Figure 9 (1-1) to (1-3)).

[0082] Furthermore, when displaying each person's opinion, the generation device 100 (output unit) can display an avatar that reproduces each person's face to the user (Figure 9 (1-4) to (1-6)). Specifically, the generation device 100 (generation unit) generates virtual emotional information of the target person as the target person's opinion. Then, based on the generated emotional information of the target person, the generation device 100 (output unit) changes the emotional display of the avatar displayed to the user.

[0083] For example, the generator 100 (generation unit) changes the avatar's emotional expression to "joy" using virtual emotional information "joy" generated in response to the mother's (woman in her 60s) opinion, "It goes well with my specialty dish," and a score of "8 points." The generator 100 (generation unit) also changes the avatar's emotional expression to "joy" using virtual emotional information "joy" generated in response to the father's (man in his 60s) opinion, "I like this!" and a score of "9 points." The generator 100 (generation unit) also changes the avatar's emotional expression to "sadness" using virtual emotional information "sadness" generated in response to the 20-year-old man's (virtual younger brother) opinion, "I'd prefer something more substantial," and a score of "2 points."

[0084] In this context, "virtual emotional information" refers to information about human emotions generated based on publicly known technologies, and includes, for example, information classified as "joy," "anger," "sadness," "happiness," "love," "hate," etc.

[0085] (Processing procedure by the generating device 100) Next, the processing procedure implemented by the generation device 100 according to this embodiment will be explained using Figure 10. Figure 10 is a flowchart showing the processing performed by the generation device 100 according to this embodiment.

[0086] The learning unit 132 sets up a profile of the person based on the person information relating to the target person (S101). If the user does not select a product (No. in S102), the generation device 100 waits for processing to begin.

[0087] If the user selects a product (Yes in S102), the generation unit 134 inputs product-related information about the selected product, user status information, and a prompt including an opinion generation command into a large-scale language model with a defined person profile (S103). The generation unit 134 then generates the opinion of the person in question (S104).

[0088] If opinions are to be aggregated (Yes in S105), the generation unit 134 aggregates the opinions of each individual whose opinions are to be aggregated (S106). On the other hand, if opinions are not to be aggregated (No in S105), the generation unit 134 skips the process in S106.

[0089] The output unit 135 outputs the generated opinion (S107). Then, the generation device 100 terminates processing.

[0090] (effect) Next, we will explain the effects of the generation device 100 according to this embodiment. Conventionally, there are known technologies for stimulating users' purchasing intent on websites, but these do not take into account the circumstances under which users select products, the profile of the person to whom they are giving the product as a gift, or their relationship with the person, and therefore there are challenges in appropriately and efficiently selecting products for purchase.

[0091] Therefore, the generation unit 134 of the generation device 100 according to this embodiment receives prompts expressed in natural language text, which include product-related information of a product selected by a user who wishes to purchase a product, user situation information, and a command to generate the target person's opinion on the product, to a large-scale language model in which the person information of the target person is set, and generates the target person's opinion on the product. The output unit 135 of the generation device 100 outputs the target person's opinion on the product generated by the generation unit 134.

[0092] Through the process described above, the generating device 100 enables users to check in advance the opinions, satisfaction level (score), etc., of a virtual person, whose profile is set to represent the target person, regarding the product they intend to purchase when they make a purchase related to a predetermined action such as giving a gift to the target person. In other words, the generating device 100 enables users to appropriately select a product that suits the target person, even when the target person's preferences are unknown or when it is difficult to confirm the target person's opinion in advance due to a surprise or other reason.

[0093] Therefore, the generation device 100 has the effect of enabling users to consider appropriate product purchases. Furthermore, the generation device 100 according to this embodiment achieves predetermined effects by performing the processes described below.

[0094] The learning unit 132 uses at least one of the following as the target person's information to set up a profile of the target person in the large-scale language model: the target person's gender and age, the relationship between the target person and the user, the results of the target person's questionnaire responses, the target person's purchasing information, and attribute information of a person similar to the target person's profile. The generation unit 134 inputs a prompt containing product-related information selected by the user, including at least one of the following: product name, price, product description, and product image, into the large-scale language model with the target person's profile set up, and generates the target person's opinion according to the relationship between the user and the target person and the purchasing situation.

[0095] Through the process described above, the generation device 100 can generate opinions from a person based on the user's own attribute information, the situation in which the user selects a product, the profile of the person the user assumes to be the target of the product selection, and the relationship between that person and the user. For example, the generation device 100 can generate opinions that the target person is expected to have, depending on whether the user is "buying a gift for a family member" or "buying a gift for a boss at work," according to the respective TPO (Time, Place, Occasion) and the target person's hobbies and preferences.

[0096] As a result, the generation device 100 has the effect of enabling users to consider appropriate product purchases by referring to opinions from a virtual person in advance when selecting products.

[0097] The generation unit 134 generates scores and opinions for each individual regarding the product, representing the opinions of the target individuals. The output unit 135 outputs the scores and opinions generated by the generation unit 134 as the opinions of the target individuals.

[0098] Through the process described above, the generator 100 can present the user with a score assigned by a virtual character for the product selected by the user, as well as the user's expected reaction when receiving the product as a gift. As a result, the generator 100 enables the user to select a product after receiving an opinion from a virtual character beforehand. Therefore, the generator 100 has the effect of enabling the user to consider purchasing products appropriately.

[0099] The generation unit 134 generates scores and opinions for each target person regarding the product, representing the opinions of the target person. Next, the generation unit 134 inputs the generated scores and opinions for multiple target people, along with a prompt containing a command to aggregate the scores and opinions of multiple target people, into a large-scale language model to generate aggregated scores and opinions. The output unit 135 outputs the aggregated scores and opinions generated by the generation unit 134 as the opinions of the target person.

[0100] Through the process described above, the generation device 100 can appropriately aggregate the opinions of multiple virtual individuals and present them to the user, even when multiple virtual individuals exist. As a result, the generation device 100 enables the user to select a product after hearing opinions from multiple virtual individuals in advance. Therefore, the generation device 100 has the effect of enabling the user to consider appropriate product purchases.

[0101] The generation unit 134 generates virtual emotional information for the target person as the target person's opinion. The output unit 135 changes the emotional display of the avatar displayed to the user based on the generated emotional information for the target person.

[0102] Through the process described above, the generator 100 changes the facial expression of a virtual person's avatar based on the opinions generated when the user selects a product, thereby enabling the user to intuitively understand whether or not the virtual person likes the product they selected. As a result, the generator 100 enables the user to consider purchasing the product appropriately.

[0103] The reception unit 133 receives opinions from real people about the product. The learning unit 132 uses the opinions of real people received by the reception unit 133 to train a large-scale language model. Through the above process, the generation device 100 enables the large-scale language model, which has a defined profile of the target person, to generate opinions that are close to those of real people when generating opinions as a virtual person. As a result, the generation device 100 enables users to obtain more appropriate opinions when selecting a product, thereby enabling users to consider purchasing the product appropriately.

[0104] <Variation> The following describes modifications that can be implemented by the generation apparatus 100 according to this embodiment.

[0105] (Data, etc.) The user information, the person profile information, the product information, the user's situation information, the subject's opinions, the aggregated opinions, the feedback, the score, the avatar, the name of the functional part of the generation device 100, the steps, the process, the name of the step or process, etc., used in the description of the above embodiment are merely examples and can be changed at will.

[0106] For example, while it was explained that User Information DB121 stores information related to each of the following items in a table format, such as "No," which is information that identifies individual data included in user information, "User Identification Information," "User Attributes," "Relationships," and "Reason for Purchase," the items stored and the information within each item are not limited. While it was explained that Person Information DB122 stores information related to each of the following items in a table format, such as "No," which is information that identifies individual data included in person information, "Person Identification Information," "Corresponding User," "Person Attributes," "Survey Information," and "Purchase Information," the items stored and the information within each item are not limited. While it was explained that Product-Related Information DB123 stores information related to each of the following items in a table format, such as "No," which is information that identifies individual data included in product-related information, "Product Name," "Product Description," "Price," and "Image," the items stored and the information within each item are not limited.

[0107] (Regarding the timing of learning) In this embodiment, it has been explained that the learning unit 132 of the generation device 100 pre-sets the profile of the target person for the large-scale language model, but the timing of learning is not limited. For example, in addition to the timing described above, the generation device 100 may include a "command to set a profile" in the prompt for the large-scale language model to execute the opinion generation process, and set the profile for the large-scale language model at the same time as the execution of the opinion generation process.

[0108] (Regarding the use of generative models) In this embodiment, the model (large-scale language model) used by the generation device 100 is described as being stored in the generation model DB 124 of the storage unit 120, but this is not limited to this. For example, the generation device 100 can access an external information processing device (server, etc.) and use a predetermined model.

[0109] (Regarding the use of large-scale language models) In this embodiment, the generation device 100 performs "setting a profile of the target person and generating opinions based on that profile" and "aggregating the generated opinions" by inputting predetermined prompts to the large-scale language model.

[0110] For example, when the generation device 100 sets up multiple person profiles in a large-scale language model, it may set up multiple person profiles in a single large-scale language model and generate opinions for each of those multiple person profiles, or it may set up the person profile for each person in a dedicated large-scale language model prepared for each person. Furthermore, the generation device 100 may implement the processing by inputting prompts corresponding to each process into a single large-scale language model for the "large-scale language model for setting person profiles" and the "large-scale language model for aggregating the generated multiple opinions," or it may implement the processing using a dedicated large-scale language model prepared for each process.

[0111] (Other examples of how user information or personal information can be used) In this embodiment, the generation device 100 is described as being able to generate opinions of a target person based on a large-scale language model in which the person's profile is defined, using user information and person information. Furthermore, the generation device 100 can generate opinions using "attributes" included in user information and "purchase history" included in person information, etc.

[0112] For example, the generator 100 can generate opinions that take into account the user's own tastes and preferences when generating opinions for a target person by including a command in the prompt that generates an opinion that reflects the user's tastes and preferences according to the "attributes" contained in the user information. In other words, the generator 100 can generate opinions that take into account not only the values ​​of the target person, but also both the user's own preferences and the values ​​of the target person.

[0113] Furthermore, the generation device 100 includes a prompt that instructs the generation of an opinion that reflects the "purchase history" included in the person's information. This allows the target person to generate an opinion that takes into account their purchase history in addition to the information from the survey. In other words, the generation device 100 can generate an opinion that considers not only the target person's values ​​based on the survey results, but also their values ​​based on their unconscious purchasing behavior.

[0114] (Product recommendations, etc.) In this embodiment, the generation device 100 can recommend products to the user using the opinions of each individual that it has generated. For example, the generation device 100 (generation unit) inputs a prompt into a large-scale language model that includes a command to select the product with the highest total score, using the "scores" of each product generated as opinions of each individual, as the recommended product. The generation device 100 (output unit) can then recommend the selected recommended product to the user.

[0115] As a result, the generating device 100 enables more efficient and effective product selection by recommending pre-calculated recommended products to the user.

[0116] (Flowcharts, etc.) In flowcharts, each step may be rearranged as long as it does not create inconsistencies, and some steps may be omitted. Furthermore, conjunctions such as "next," "continue," "in addition," "at this time," and "on this occasion" in flowchart descriptions do not limit the order or timing of the processes in the flowchart.

[0117] <Hardware Configuration> Each component of the illustrated device is a functional concept and does not necessarily have to be physically configured as shown. In other words, the specific forms of distribution and integration of each device are not limited to those shown, and all or part of them can be functionally or physically distributed and integrated in any unit according to various loads and usage conditions. Furthermore, each processing function performed by each device can be implemented, all or any part of it, by a CPU and the program that is analyzed and executed by that CPU, or by hardware using wired logic.

[0118] Furthermore, among the processes described in this embodiment, all or part of those described as being performed automatically can be performed manually using known methods. In addition, the processing procedures, control procedures, specific names, and information including various data and parameters shown in the drawings can be arbitrarily changed unless otherwise specified.

[0119] <Program> In one embodiment, the various devices constituting the generation device 100 can be implemented by installing the generation program as packaged software or online software on a desired computer. For example, by having the above generation program executed on an information processing device, the various devices constituting the generation device 100 can be made to function. The information processing device referred to here includes desktop or notebook personal computers. In addition, the information processing device also includes mobile communication terminals such as smartphones and mobile phones, and slate terminals such as PDAs (Personal Digital Assistants).

[0120] Figure 11 shows an example of a computer that implements the generation device 100 according to the embodiment. The computer 1000 has, for example, memory 1010 and CPU 1020. The computer 1000 also has a hard disk drive interface 1030, a disk drive interface 1040, a serial port interface 1050, a video adapter 1060, and a network interface 1070. These components are connected by a bus 1080.

[0121] Memory 1010 includes ROM (Read Only Memory) 1011 and RAM 1012. ROM 1011 stores, for example, a boot program such as BIOS (Basic Input Output System). The hard disk drive interface 1030 is connected to the hard disk drive 1090. The disk drive interface 1040 is connected to the disk drive 1100. For example, a removable storage medium such as a magnetic disk or optical disk is inserted into the disk drive 1100. The serial port interface 1050 is connected to, for example, a mouse 1110 and a keyboard 1120. The video adapter 1060 is connected to, for example, a display 1130.

[0122] The hard disk drive 1090 stores, for example, an OS (Operating System) 1091, an application program 1092, a program module 1093, and program data 1094. That is, the programs that define the various processes of the various devices constituting the generation device 100 are implemented as program modules 1093 in which executable code for a computer is written. The program modules 1093 are stored, for example, in the hard disk drive 1090. For example, a program module 1093 for performing processes similar to the functional configuration of the various devices constituting the generation device 100 is stored in the hard disk drive 1090. Note that the hard disk drive 1090 may be replaced by an SSD (Solid State Drive).

[0123] Furthermore, the configuration data used in the processing of the embodiment described above is stored as program data 1094 in, for example, memory 1010 or hard disk drive 1090. The CPU 1020 then reads the program module 1093 and program data 1094 stored in memory 1010 or hard disk drive 1090 into RAM 1012 as needed and executes the processing of the embodiment described above.

[0124] Furthermore, the program module 1093 and program data 1094 are not limited to being stored in the hard disk drive 1090; for example, they may be stored in a removable storage medium and read by the CPU 1020 via a disk drive 1100 or the like. Alternatively, the program module 1093 and program data 1094 may be stored in another computer connected via a network (LAN, WAN (Wide Area Network), etc.). The program module 1093 and program data 1094 may then be read from the other computer by the CPU 1020 via a network interface 1070.

[0125] <Other> Although this embodiment has been described above, this embodiment is not limited by the description and drawings that constitute part of the disclosure. That is, all other embodiments, examples, and operational techniques made by those skilled in the art based on this embodiment are included in the scope of this embodiment. [Explanation of symbols]

[0126] 100 generator 110 Communications Department 120 Storage section 121 User Information Database 122 Person information DB 123 Product-related information database 124 Generative Model DB 130 Control Unit 131 Acquisition Department 132 Learning Department 133 Reception Department 134 Generation part 135 Output section

Claims

1. For a large-scale language model in which information about the target person's profile has been set, A generation unit receives prompts expressed in natural language text, which include information about a product selected by a user who wishes to purchase the product, information about the user's circumstances, and a command to generate the opinion of the person concerned regarding the product, and generates the opinion of the person concerned regarding the product. An output unit that outputs the opinion of the person concerned regarding the product generated by the generation unit, A generating apparatus characterized by having the following features.

2. The system further includes a learning unit that sets up a profile of the target person for the large-scale language model using at least one of the following as information regarding the target person's profile: the target person's gender and age, the relationship between the target person and the user, the results of a questionnaire given by the target person, the target person's purchasing information, and attribute information of a person similar to the target person's profile. The generating unit is The prompt, which includes information about the product selected by the user, including at least one of the product name, price, product description, and product image, is input to the large-scale language model in which the profile of the target person is set. The system generates the opinions of the target person based on the relationship between the user and the target person and their purchasing status. The generating apparatus according to feature 1.

3. The system further includes a reception area for receiving opinions from real people regarding the aforementioned product. The aforementioned learning unit, The large-scale language model is trained using the opinions of the real person received by the reception department. The generating apparatus according to feature 2.

4. The generating unit is As the opinions of the aforementioned individuals, a score and impression of the aforementioned product are generated for each individual. The output unit is, The score and comments generated by the generation unit are output as the opinion of the person in question. The generating apparatus according to any one of claims 1 to 3.

5. The generating unit is As the opinions of the aforementioned individuals, a score and comments on the aforementioned product are generated for each individual, The scores and comments for multiple target individuals that have been generated, and the prompts including a command to aggregate the scores and comments for multiple target individuals, are input to the large-scale language model to generate the aggregated scores and comments. The output unit is, The aggregated scores and comments generated by the generation unit are output as the opinion of the person concerned. The generating apparatus according to any one of claims 1 to 3.

6. The generating unit is The virtual emotional information of the subject person is generated as the opinion of the subject person. The output unit is, Based on the generated emotional information of the target person, the emotional display of the avatar shown to the user is changed. The generating apparatus according to any one of claims 1 to 3.

7. A generation method to be executed by a generation device, For a large-scale language model in which information about the target person's profile has been set, A generation process that generates an opinion of a person regarding a product by inputting prompts expressed in natural language text, which include information about a product selected by a user who wishes to purchase the product, information about the circumstances of the user who wishes to purchase the product, and a command to generate an opinion of the person regarding the product. An output step which outputs the opinion of the target person regarding the product generated by the generation step, A method for generating a product, characterized by including the following:

8. For a large-scale language model in which information about the target person's profile has been set, A generation step in which a prompt is input, which is expressed in natural language text, containing information about a product selected by a user who wishes to purchase the product, information about the circumstances of the user who wishes to purchase the product, and a command to generate the opinion of the person concerned regarding the product, and generates the opinion of the person concerned regarding the product. An output step which outputs the opinion of the person concerned regarding the product generated in the generation step, A generation program that causes a computer to execute something.

Images