Mediation device, method for mediation, and computer program

Abstract

To suppress insertion of an unnatural advertisement into a conversation between a user and an AI.SOLUTION: A mediation device for mediating a conversation between a user and an AI sends at least one topic as topic information to the AI. The device selects a product of an appeal target related to a designated conversation, which is a conversation including one of the at least one topic, from the at least one product on the condition that the AI notifies that a conversation with the user is the designated conversation on the basis of the topic information. The device executes insertion processing of inserting the appeal information into a conversation to appeal the selected appeal target product to the user.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] The present disclosure relates to an intermediary device. [Background technology]

[0002] 2. Description of the Related Art There is known an artificial intelligence (AI) that responds to text or voice input by a user and converses with the user. For example, Patent Document 1 discloses a system that acquires data from various social networking services used by users, stores the data in a database, and uses artificial intelligence to learn the user's thoughts and provide appropriate answers to questions from the user. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 6831522 Summary of the Invention [Problem to be solved by the invention]

[0004] The present inventors are considering inserting advertisements into conversations between users and artificial intelligence. However, if the advertisement to be inserted has low relevance to the conversation context, the advertisement may be inserted abruptly and unnaturally.

[0005] One aspect of the present disclosure is to suppress unnatural advertisement insertion in conversations between users and artificial intelligence. [Means for solving the problem]

[0006] One aspect of the present disclosure is an intermediary device that mediates a conversation between a user and an artificial intelligence, comprising a receiving unit and a control unit. The receiving unit is configured to receive an input sentence from a user input via a platform. The control unit is configured to transfer the input sentence received by the receiving unit to the artificial intelligence, and to output a generated sentence generated by the artificial intelligence as a response to the input sentence to the platform.

[0007] The control unit further transmits one or more topics to the artificial intelligence as topic information. Based on the topic information, on condition that the artificial intelligence notifies the control unit that the conversation with the user is a designated conversation that includes one or more topics, the control unit selects, from one or more products, a promotion target product related to the designated conversation. The control unit executes an insertion process for inserting promotion information for promoting the selected promotion target product to the user into the conversation.

[0008] With this configuration, unnatural advertisement insertion in conversations between users and artificial intelligence can be suppressed. In one aspect of the present disclosure, the insertion process may be a process of instructing an artificial intelligence to generate a sentence including appeal information as a response to an input sentence.

[0009] In one aspect of the present disclosure, when the control unit obtains a sentence including appeal information as a generated sentence from the artificial intelligence, the control unit may add information including at least one of a description, an image, a link to a website, and an advertising creative related to the promoted product to the sentence including the appeal information and output the same to the platform.

[0010] With this configuration, additional appeal information can be added to the generated sentence that includes appeal information obtained from the artificial intelligence. In one aspect of the disclosure, the control unit may be configured to instruct the artificial intelligence to steer the conversation towards any of one or more topics.

[0011] According to this configuration, it is possible to increase the possibility that product appeal information can be naturally inserted into a conversation as an advertisement. In one aspect of the present disclosure, the prompt may be one to which the user can respond with one of two responses: a positive response and a negative response.

[0012] With this configuration, it becomes easy to guide the conversation from the topic to be guided to the product to be promoted. In one aspect of the present disclosure, the control unit may select one or more topics from a plurality of weighted topics with a probability corresponding to the weighting, and transmit the selected one or more topics to the artificial intelligence as topic information.

[0013] With this configuration, the probability of selecting a product that is more desired to be promoted as the product to be promoted can be increased. In one aspect of the present disclosure, the control unit may select a product to be promoted based on a keyword included in the specified conversation.

[0014] In one embodiment of the present disclosure, each of the one or more products may be associated with a keyword. The control unit may select, from among the one or more products, a product associated with a keyword included in the specified conversation as a product to be promoted.

[0015] In one embodiment of the present disclosure, one or more topics may each be associated with one or more products. The control unit may select a product to be promoted from one or more products associated with the topics included in the designated conversation on condition that the artificial intelligence notifies the control unit that the conversation is a designated conversation.

[0016] In one aspect of the present disclosure, the control unit may acquire attribute information on the attributes of a user who converses with the artificial intelligence from at least one of the artificial intelligence and the platform. The control unit may also select a product to be promoted by taking into account the attributes of the user identified from the acquired attribute information.

[0017] According to this configuration, a product that is more appealing to the user can be selected as the product to be promoted. In one aspect of the present disclosure, the control unit may transmit one or more first topics to the artificial intelligence as first topic information. The control unit may also transmit one or more second topics related to the topic included in the first designated conversation as second topic information to the artificial intelligence on the condition that the artificial intelligence notifies the user that the conversation with the user is a first designated conversation that is a conversation that includes one or more first topics based on the first topic information. The control unit may also select a product to be promoted related to the second designated conversation from among one or more products on the condition that the artificial intelligence notifies the user that the conversation with the user is a second designated conversation that is a conversation that includes one or more second topics based on the second topic information.

[0018] With this configuration, advertisement insertion based on topic transitions can be realized, and unnatural advertisement insertion in conversations between users and artificial intelligence can be suppressed. In one aspect of the present disclosure, the control unit may select a product to be promoted based on a keyword included in the second specified conversation.

[0019] In one embodiment of the present disclosure, each of the one or more products may be associated with a keyword. The control unit may select, from among the one or more products, a product associated with a keyword included in the second specified conversation as a product to be promoted.

[0020] In one aspect of the present disclosure, the control unit may select a first guidance target topic from among the multiple topics, and transmit the first guidance target topic to the artificial intelligence together with information instructing the artificial intelligence to guide the conversation to the topic of the first guidance target. After transmitting the first guidance target topic, the control unit may determine whether the artificial intelligence has failed to guide the conversation to the topic of the first guidance target. If the control unit determines that the artificial intelligence has failed, it may select a topic different from the first guidance target topic from among the multiple topics as a second guidance target topic, and transmit the second guidance target topic to the artificial intelligence together with information instructing the artificial intelligence to guide the conversation to the topic of the second guidance target.

[0021] According to this configuration, it is possible to increase the possibility that product appeal information can be naturally inserted into a conversation as an advertisement. In one embodiment of the present disclosure, the device may further include a storage unit configured to hold a table associating one or more products related to each of the one or more topics. The control unit may select the products to be promoted based on the table.

[0022] With this configuration, unnatural advertisement insertion in conversations between users and artificial intelligence can be suppressed. In one aspect of the present disclosure, a plurality of topics may be registered in the table, and the control unit may select one or more topics from the plurality of topics registered in the table, and transmit the selected one or more topics as topic information.

[0023] One aspect of the present disclosure is an intermediary device that mediates a conversation between a user and an artificial intelligence, comprising a receiving unit and a control unit. The receiving unit receives an input sentence from a user input via a platform. The control unit is configured to transfer the input sentence received by the receiving unit to the artificial intelligence, and output a generated sentence generated by the artificial intelligence as a response to the input sentence to the platform. The control unit further transmits one or more topics to the artificial intelligence as topic information. When a notification is received from the artificial intelligence based on the topic information indicating that a conversation with a user has become one of one or more topics, a promotion target product related to the topic corresponding to the notification is selected from one or more products. An insertion process is performed to insert promotion information for promoting the selected promotion target product to the user into the conversation.

[0024] In one aspect of the present disclosure, an intermediation method corresponding to the intermediation device described above may be provided. In one aspect of the present disclosure, a mediation method for mediating a conversation between a user and an artificial intelligence, which is executed by a computer, may be provided. The mediation method includes receiving an input sentence from a user inputted through a platform, transferring the received input sentence to the artificial intelligence, outputting a generated sentence generated by the artificial intelligence as a response to the input sentence to the platform, transmitting one or more topics to the artificial intelligence as topic information, selecting a product to be promoted related to the designated conversation from one or more products based on the topic information, on the condition that the artificial intelligence notifies the user that the conversation with the user is a designated conversation including one or more topics, and performing an insertion process for inserting promotion information for promoting the selected product to the user into the conversation. This mediation method has the same effect as the mediation device described above.

[0025] In one aspect of the present disclosure, a computer program for causing a computer to execute the mediation method may be provided. The computer program may be recorded in a computer-readable recording medium. [Brief description of the drawings]

[0026] [Figure 1] FIG. 1 is a block diagram showing a configuration of a chat system. [Diagram 2] FIG. 4 is a diagram showing an example of a topic list and a product list in the first embodiment. [Diagram 3] 4 is a flowchart of an intermediation process in the first embodiment. [Figure 4] FIG. 11 is a diagram showing an example of a table in the second embodiment. [Diagram 5] 13 is a flowchart of an intermediation process when a topic is presented by a user in the second embodiment. [Figure 6] A figure showing an example of a conversation between a user and artificial intelligence and the processing of the intermediary device when a topic is presented by the user in the second embodiment. [Figure 7]13 is a flowchart of an intermediation process in the second embodiment when a topic is presented by artificial intelligence. [Figure 8] A figure showing an example of a conversation between a user and an artificial intelligence and the processing of the intermediary device when a topic is suggested by the artificial intelligence in the second embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0027] Hereinafter, exemplary embodiments of the present disclosure will be described with reference to the drawings. [1. First embodiment] [1-1. Configuration] The conversation system 1 shown in Figure 1 is a system in which a user has a conversation with an artificial intelligence 5, which responds to text, voice, etc. input by the user to a user terminal 2 via a platform 3 and an intermediary device 4.

[0028] The conversation system 1 includes a user terminal 2, a platform 3, an intermediary device 4, and an artificial intelligence 5. The user terminal 2 is configured to be able to communicate with the platform 3. The platform 3 is configured to be able to communicate with the user terminal 2 and the intermediary device 4. The intermediary device 4 is configured to be able to communicate with the platform 3 and the artificial intelligence 5. At least one of the user terminal 2, the platform 3, the intermediary device 4, and the artificial intelligence 5 may be configured to operate on a cloud network.

[0029] The user terminal 2 is a terminal that accepts input of text, voice, etc. from a user. Examples of the user terminal 2 include a personal computer and a smartphone. The user can input text, voice, and the like via a keyboard, touch panel, microphone, and the like provided in the user terminal 2. Hereinafter, the text input by the user to the user terminal 2 (in the case of voice input, the voice converted into text) is referred to as an input sentence. The input sentence is transmitted from the user terminal 2 to the platform 3.

[0030] The platform 3 is a server that cooperates with applications installed in the user terminal 2 to realize message delivery and chatbot functions. An example of the platform 3 is the LINE platform. LINE is a registered trademark.

[0031] The platform 3 transmits an input sentence received from the user terminal 2 to the intermediary device 4, and receives a text (hereinafter referred to as a generated sentence) generated by an artificial intelligence 5 as a response to the input sentence from the intermediary device 4. The input sentence and the generated sentence are displayed on an application installed in the user terminal 2 via the platform 3.

[0032] The intermediary device 4 is a device that mediates conversation between a user and an artificial intelligence 5. The intermediary device 4 transmits an input sentence received from the platform 3 to the artificial intelligence 5, and transmits a generated sentence received from the artificial intelligence 5 to the platform 3. The generated sentence transmitted to the platform 3 is displayed on the user terminal 2.

[0033] The intermediary device 4 can add any content when transmitting an input sentence to the artificial intelligence 5 or a generated sentence to the platform 3. For example, when transmitting an input sentence to the artificial intelligence 5, the intermediary device 4 can also transmit an instruction sentence for the artificial intelligence 5 to the artificial intelligence 5.

[0034] For example, when the intermediary device 4 transmits an input sentence to the artificial intelligence 5, it can additionally instruct the artificial intelligence 5 what kind of generated sentence it wants the artificial intelligence 5 to generate in response to the input sentence by transmitting an instruction sentence. Alternatively, when the intermediary device 4 transmits a generated sentence to the platform 3, it can add information for promoting a specific product (hereinafter, referred to as appeal information) to the generated sentence.

[0035] As an example, the intermediary device 4 can instruct the artificial intelligence 5 to "while talking to the user, naturally draw out the user's concerns about 'physical pain.'" Alternatively, when the generated text contains a character string such as a product name representing the product being appealed for, the intermediary device 4 can transmit the text to the platform 3 with an added description of the corresponding product, an image of the product, a link to a website introducing the product, and the like. Products being appealed for include both tangible goods and intangible services.

[0036] The intermediary device 4 includes a processor 41, a memory 42, a storage 43, and a communication interface 44. The processor 41 is configured to execute processing in accordance with a computer program recorded in the storage 43 .

[0037] The memory 42 is used as a working area when the processor 41 executes a process. An example of the memory 42 is a RAM. The storage 43 holds various data as well as computer programs executed by the processor 41. Examples of the storage 43 include a hard disk drive (HDD) and a solid state drive (SSD).

[0038] The various data held by storage 43 includes list 6, which includes one or more topics related to one or more products that are the subject of promotion. Fig. 2 shows an example of list 6, which includes at least the topics "cooking," "health," "exercise," "love," and "games."

[0039] The communication interface 44 is configured to be able to communicate with an external system through a wide area network such as the Internet. The intermediary device 4 is configured to be able to communicate with the platform 3 and the artificial intelligence 5 through the communication interface 44.

[0040] The artificial intelligence 5 is a chatbot that is constructed by machine learning such as deep learning based on a large language model (LLM) and can have a natural conversation with a user. An example of the artificial intelligence 5 is ChatGPT (internationally registered trademark) by OpenAI. When an input sentence is input from the intermediary device 4, the artificial intelligence 5 generates a generated sentence as a response to the input sentence and transmits it to the intermediary device 4. When an instruction is input from the intermediary device 4 in addition to the input sentence, the artificial intelligence 5 generates a generated sentence based on the instruction.

[0041] [1-2. Processing] The intermediation process executed by the processor 41 will be described with reference to the flowchart of FIG. When the communication interface 44 receives an input sentence transmitted from the user via the user terminal 2 and the platform 3, the processor 41 assumes that a conversation has started and starts the mediation process shown in FIG.

[0042] First, in S100, the processor 41 transmits the contents of the list 6 to the artificial intelligence 5. The process of S100 may be executed before the conversation is started. Next, in S105, the processor 41 acquires the input sentence received by the communication interface 44.

[0043] Next, at S110, the processor 41 instructs the artificial intelligence 5 to notify via the communication interface 44 whether the input sentence is an input sentence that includes any one of the topics included in the list 6 (hereinafter referred to as a designated conversation). The processor 41 instructs the artificial intelligence 5 to include in the notification information that can identify which of the topics included in the list 6 the conversation has become. Upon receiving the instruction, if the input sentence is a designated conversation, the artificial intelligence 5 notifies the intermediary device 4 that the input sentence is a designated conversation.

[0044] When the list 6 transmitted from the processor 41 to the artificial intelligence 5 contains one topic, the processor 41 may instruct the artificial intelligence 5 to include in the notification, instead of information that can identify the topic of the conversation, information that can determine that the topic of the input sentence has become a topic included in the list 6 transmitted to the artificial intelligence 5. When the intermediary device 4 receives a notification that includes such information, it identifies the topic of the conversation based on the list 6 stored in the storage 43.

[0045] As an example, the processor 41 can send an input sentence to the artificial intelligence 5 and instruct the artificial intelligence 5 to identify whether the topic related to the input sentence is a topic included in the list 6. For example, when the input sentence is "I haven't been getting enough sleep lately," the processor 41 can instruct the artificial intelligence 5 to "identify whether the topic of the next input sentence is one of "cooking," "health," "exercise," "romance," and "games." If identified, reply "The input sentence is a specified conversation. Input sentence: "I haven't been getting enough sleep lately." In response to this instruction, the artificial intelligence 5 can identify the topic of the input sentence as "health" and notify the intermediary device 4 that the input sentence is a specified conversation.

[0046] Alternatively, when ChatGPT (international registered trademark) is used as the artificial intelligence 5, the processor 41 may instruct the artificial intelligence 5 to identify a topic related to the input sentence and to notify that the input sentence is a specified conversation by using the Function Calling function of ChatGPT. The Function Calling function is a function in which, in a state in which a function program (hereinafter, a function) is input to the artificial intelligence 5, the artificial intelligence 5 calls a function related to the input as necessary and applies the output result of the function to the text generated as a response to the input. The function includes explanatory information that explains what the function is. For example, a function including information such as "Identify the topic of the current conversation" and "Select a conversation topic from a list" as explanatory information, and parameters having a list of "cooking", "health", "exercise", "love", and "games" may be input to the artificial intelligence 5 together with the input sentence and information instructing the artificial intelligence 5 to identify the topic of the input sentence. In this case, when the artificial intelligence 5 is able to identify the topic of the input sentence, it calls the function to identify the topic of the input sentence and notifies the intermediary device 4 whether the input sentence is a specified conversation or not.

[0047] Next, in S115, the processor 41 determines, based on the notification from the artificial intelligence 5, whether the input sentence is a specified conversation. When the processor 41 determines in S115 that the input sentence is not the specified conversation (S115: NO), the processor 41 proceeds to S120. In S120, the processor 41 transmits the input sentence to the artificial intelligence 5 and instructs the artificial intelligence 5 to guide the conversation to a topic selected from the list 6 as a topic to be guided. The method of selecting a topic is not particularly limited, and for example, a topic may be selected randomly. The processor 41 can select one or more topics as a topic to be guided.

[0048] When guiding the conversation to the topic to be guided, the processor 41 may instruct the artificial intelligence 5 to guide the conversation to a topic to be guided in a way that the user can respond with either a positive or negative response. For example, the processor 41 may instruct the artificial intelligence 5 to generate a generated sentence that guides the conversation to the topic to be guided, to be a question that the user can respond with "yes" or "no". An example of the content of the instruction to the artificial intelligence 5 may be an instruction including the phrase "Please guide the conversation so that the topic of the conversation becomes "health". Please guide the conversation to be a question that can be answered with "yes" or "no". In accordance with this instruction, the artificial intelligence 5 can generate a response sentence to the input sentence as a generated sentence, including a content that can guide the topic to the topic of "health".

[0049] Next, in S125, the processor 41 receives the generated sentence from the artificial intelligence 5 and transmits it to the platform 3 via the communication interface 44, and then returns to S105. On the other hand, when the processor 41 determines in S115 that the input sentence is a specified conversation (S115: YES), the process proceeds to S130. In S130, the processor 41 selects a promotion target product related to the input sentence that is the specified conversation from among one or more promotion targets.

[0050] For product selection, the intermediary device 4 can store in the storage 43 a product list 6P that describes, for each topic listed in the list 6 shown in Fig. 2, one or more products related to the corresponding topic as a topic-related product group. The same products may be included in the product lists for different topics. As illustrated in Fig. 2, for the topic "exercise", a product list describing multiple products related to "exercise", including "tennis rackets" and "golf clubs", as the topic-related product group may be prepared.

[0051] For example, when the topic of the specified conversation is "exercise", the processor 11 can select one product to be promoted from a topic-related product group including "tennis rackets" and "golf clubs" included in a product list related to the topic "exercise". The selection method is not particularly limited, and for example, the product to be promoted may be selected randomly.

[0052] Alternatively, the products to be promoted may be selected taking into account the attributes of the user who is conversing with the artificial intelligence 5. For this purpose, the processor 41 can acquire attribute information on the attributes of the user from at least one of the intermediary device 4 and the artificial intelligence 5. The processor 41 can select the product to be promoted from the corresponding trending product list, taking into account the user's attributes identified from the acquired attribute information. Examples of the attribute information include the user's gender, age, etc.

[0053] The product list may include target attribute information for each product. The target attributes may be understood to mean attributes of consumers that are important as targets for the product. The target attributes may be expressed by the age group and gender of the corresponding consumers. The processor 41 may select the products to be promoted from the group of topic-related products based on a comparison between the user's attributes and the target attributes of each product, such that products with target attributes that are highly consistent with the user's attributes are preferentially selected as the products to be promoted.

[0054] By using attribute information, for example, when there are products for men and products for women as promotional products for the same topic, processor 41 can select the product that matches the user's gender as the promotional product.

[0055] According to the example shown in Fig. 2, "tennis rackets" are associated with "women in their 20s" as a target attribute, and "golf clubs" are associated with "men in their 40s" as a target attribute. In this case, when the user is a "man in his 40s," he can select "golf clubs," which are of high interest to "men in their 40s," as the product to be promoted from among a group of topic-related products that include "tennis rackets" and "golf clubs."

[0056] When acquiring attribute information from the artificial intelligence 5, the processor 41 may acquire the attribute information from the artificial intelligence 5, for example, by instructing the artificial intelligence 5 to estimate attribute information regarding the user's attributes based on an input sentence. When acquiring attribute information from the platform 3, the processor 41 may acquire the attribute information from account information of the user registered on the platform 3, for example.

[0057] Next, in S135, the processor 41 executes an insertion process for inserting the appeal information of the selected product to be promoted into the input text, and proceeds to S125. Examples of the appeal information include a description of the product to be promoted, an image, a link to a website, and an advertisement creative. The insertion process is, for example, a process in which the processor 41 transmits an input text to the artificial intelligence 5 and instructs the artificial intelligence 5 to generate a text including the appeal information of the selected product to be promoted as a response to the input text. Alternatively, when the processor 41 proceeds from S135 to S125 as the insertion process, the processor 41 may add other appeal information to the generated text acquired from the artificial intelligence 5 and then transmit it to the platform 3. For example, the processor 41 may transmit the input text to the artificial intelligence 5, instruct the artificial intelligence 5 to generate a text including the name of the product to be promoted in S135, and when the name of the product is included in the generated text acquired in S125, add a description of the product, an image, etc., and transmit the generated text to the platform 3.

[0058] In this manner, if the conversation topic is included in list 6, processor 41 executes a process for inserting product appeal information into the conversation. [1-3. Effects] According to the embodiment described above in detail, the following effects can be obtained.

[0059] (1a) When the input sentence is a specified conversation, the processor 41 selects a product to be promoted that is related to the topic of the input sentence, which is the specified conversation, from among one or more products to be promoted. The processor 41 transmits the input sentence to the artificial intelligence 5 and instructs the artificial intelligence 5 to generate a sentence including promotion information of the product to be promoted. The processor 41 receives the generated sentence from the artificial intelligence 5 and transmits it to the platform 3 via the communication interface 44.

[0060] According to this configuration, it is possible to promote products related to the topic of conversation rather than promoting products regardless of the topic of conversation. Therefore, it is possible to suppress unnatural advertisement insertion in the conversation between the user and the artificial intelligence 5.

[0061] (1b) If the input sentence is not a specified conversation, the processor 41 sends the input sentence to the artificial intelligence 5 and instructs the artificial intelligence 5 to guide the conversation to one of the topics selected from the list 6 as the topic to be guided.

[0062] According to this configuration, the conversation topic is more likely to be related to the product than when the topic is not steered, which increases the possibility that the product's appeal information can be naturally inserted into the conversation as an advertisement.

[0063] (1c) When guiding a conversation to a topic to be guided, the processor 41 may instruct the artificial intelligence 5 to provide guidance that the user can respond to with either a positive or negative response.

[0064] According to such a configuration, it becomes easy for the user to specify a topic in which he or she is interested, and therefore it becomes easy to lead the conversation from the topic to be led to the product to be promoted. (1d) The processor 41 acquires attribute information relating to the attributes of the user, and selects a product to be promoted, taking into consideration the attributes of the user identified from the attribute information.

[0065] According to this configuration, products related to the attributes of the user can be selected as products to be promoted. (1e) List 6 contains one or more topics related to each of the one or more products being claimed.

[0066] According to such a configuration, the more topics included in the list 6, the higher the possibility of leading the conversation to a topic similar to the previous conversation. Therefore, unnatural advertisement insertion in the conversation between the user and the artificial intelligence 5 can be further suppressed.

[0067] [1-4. Correspondence between terms] In the above embodiment, the input sentence corresponds to an example of a conversation, the communication interface 44 corresponds to an example of a receiving unit, the processor 41 corresponds to an example of a control unit, the list 6 corresponds to an example of topic information, and the product list 6P corresponds to an example of one or more products associated with each of the one or more topics.

[0068] [2. Second embodiment] [2-1. Configuration] In the second embodiment, the basic configuration of the conversation system 1 is the same as in the first embodiment. Below, configurations of the conversation system 1 of the second embodiment that are different from those of the first embodiment will be selectively described. In the second embodiment, configurations that are given the same reference numerals as in the first embodiment may be understood to be the same configurations as in the first embodiment unless otherwise described below.

[0069] The various data held by storage 43 includes table 7 that associates one or more topics related to each of one or more products to be promoted with one or more keywords related to the products. In addition to the topics and keywords related to each product, table 7 includes at least one of data items including a description, an image, a link to a website, and an advertising creative for each product. An example of a keyword is a user's concern or need that can be solved by the product to be promoted. The user's concern or need related to each keyword may be associated with each keyword in table 7.

[0070] 4 includes records corresponding to seasoning A, medicine B, and sneakers C as products to be promoted. The products to be promoted here are products that are distinguished by, for example, product names or product codes. For example, seasoning A may be one of the seasoning products that are distinguished by product names.

[0071] The record for seasoning A in Table 7 includes the topics "cooking," "home cooking," and "menu," the keywords "recipe" and "how to make"; the description "It adds an accent to dishes and brings out their flavor"; the image A of seasoning A; and the link "https: / / xxx.com," a link to a website about seasoning A.

[0072] The record of Drug B in Table 7 includes topics "health" and "body pain", keywords "lower back pain" and "stiff shoulders", description "Effective for lower back pain and stiff shoulders. Gentle on the skin. Can also be used for muscle pain and joint pain", image a of Drug B, and link "https: / / yyy.com" to a website about Drug B. The topic "body pain" may be a concern of the user.

[0073] The record for sneakers C in Table 7 includes the topics “exercise” and “sports”, the keyword “running”, the description “Lightweight, breathable, and highly shock-absorbing, putting less strain on the feet”, the image c of sneakers C, and the link “https: / / zzz.com”, a link to a website about sneakers C.

[0074] A different topic may be registered for each product, or the same topic may be registered for each product in table 7. For example, if seasoning A and seasoning D are promoted products, the topics associated with seasoning A and seasoning D may overlap in part or in whole. In other words, a single topic may be associated with multiple products in promotion information.

[0075] The contents of table 7 are input in advance to the artificial intelligence 5. Alternatively, a pointer to table 7 (in other words, the address where table 7 is recorded in the storage 43) is input. Therefore, the artificial intelligence 5 can refer to the contents of table 7 at any time.

[0076] [2-2. Processing] [2-2-1. Mediating process when a topic is suggested by the user] The intermediation process executed by the processor 41 will be described with reference to the flowchart of FIG.

[0077] When the communication interface 44 receives an input sentence transmitted from the user via the user terminal 2 and the platform 3, the processor 41 starts the intermediation process shown in FIG. First, in S200, the processor 41 acquires an input sentence received by the communication interface 44.

[0078] Next, in S205, the processor 41 instructs the artificial intelligence 5 to notify via the communication interface 44 whether the input sentence is an input sentence that includes any of the topics included in the table 7 (hereinafter, referred to as a designated conversation). If the input sentence is a designated conversation, the artificial intelligence 5 that has received the instruction notifies the intermediary device 4 that the input sentence is a designated conversation.

[0079] Next, in S210, the processor 41 determines whether or not the input sentence is a specified conversation based on the notification from the artificial intelligence 5. Note that the processor 41 may instruct the artificial intelligence 5 to identify a topic of the input sentence in S205, and then determine whether or not the identified topic is included in the table 7 in S210.

[0080] When the processor 41 determines in S210 that the input sentence is not the specified conversation (S210: NO), the processor 41 proceeds to S215. In S215, the processor 41 transmits the input sentence to the artificial intelligence 5 and instructs the artificial intelligence 5 to guide the conversation to a topic selected from the table 7 as a topic to be guided. The method of selecting a topic is not particularly limited, and for example, the processor 41 may select a topic randomly. The processor 41 can select one or more topics as a topic to be guided.

[0081] When guiding the conversation to a topic to be guided, the processor 41 may instruct the artificial intelligence 5 to guide the conversation to a topic to which the user can respond with either a positive or negative response. For example, the processor 41 may instruct the artificial intelligence 5 to generate a generated sentence that guides the conversation to the topic to be guided, such that the generated sentence is a question that the user can respond with "yes" or "no." An example of the instructions to be given to the artificial intelligence 5 may be an instruction including the phrase "Please guide the conversation to the topic of 'health.' Please guide the conversation in a question that can be answered with "yes" or "no."

[0082] Next, in S220, the processor 41 receives the generated sentence from the artificial intelligence 5 and transmits it to the platform 3 via the communication interface 44, and then returns to S200. On the other hand, if the processor 41 determines in S210 that the input sentence is a specified conversation (S210: YES), the processor 41 proceeds to S225 and determines whether or not a keyword corresponding to the topic of the specified conversation is included in the input sentence.

[0083] When the processor 41 determines in S225 that the input sentence does not include a keyword corresponding to the topic of the specified conversation (S225: NO), the processor 41 proceeds to S230. In S230, the processor 41 transmits the input sentence to the artificial intelligence 5 and instructs the artificial intelligence 5 to guide the artificial intelligence 5 to a topic that elicits one or more keywords corresponding to the topic of the specified conversation. As an example of the content of the instruction to the artificial intelligence 5, a fixed phrase corresponding to each keyword can be stored in the table 7 in advance. The processor 41 can instruct the artificial intelligence 5 to use the corresponding fixed phrase to guide the artificial intelligence 5 to a topic that elicits the keyword. The fixed phrase includes information that instructs the artificial intelligence 5 to ask a specific question to elicit the corresponding keyword. For example, as a fixed phrase to elicit the keyword "lower back pain" or "stiff shoulders", the instruction content "Please talk naturally to elicit whether there is any part of your body that hurts" can be stored in the table 7. When the processor 41 instructs the artificial intelligence 5 to guide the artificial intelligence 5 to a topic that elicits the keyword "lower back pain" or "stiff shoulders" in S230, the processor 41 can instruct the artificial intelligence 5 using this fixed phrase.

[0084] When extracting a keyword corresponding to the designated conversation topic, the processor 41 may instruct the artificial intelligence 5 to be able to specifically answer the user's concerns and needs. For example, to extract the keyword "lower back pain", the processor 41 may instruct to inquire about the user's concern "body pain" related to the keyword "lower back pain". An example of an expected generated sentence is "Does your body hurt anywhere?". After that, the processor 41 proceeds to S220.

[0085] On the other hand, when the processor 41 determines in S225 that the input sentence contains a keyword corresponding to the designated conversation topic (S225: YES), the processor 41 proceeds to S235. In S235, the processor 41 selects a product associated with the keyword in the table 7 as a product to be promoted from among one or more products to be promoted.

[0086] At this time, the processor 41 may acquire attribute information on the attributes of the user, and select the product to be promoted by taking into account the user's attributes identified from the attribute information. Examples of attribute information include the gender and age of the user. For example, when target attribute information, which is information on the target attribute, is associated with each of the products to be promoted in the table 7, the processor 41 selects a product that matches the target attribute information associated with the product corresponding to the keyword as the product to be promoted. As an example, when the sneakers C in the table 7 are divided into men's sneakers C(M) whose target attribute information is for men and women's sneakers C(F) whose target attribute information is for women, the processor 41 determines whether to select the men's sneakers C(M) or the women's sneakers C(F) as the product to be promoted based on the gender of the user in the attribute information.

[0087] When acquiring attribute information from the artificial intelligence 5, the processor 41 may acquire the attribute information from the artificial intelligence 5, for example, by instructing the artificial intelligence 5 to estimate attribute information regarding the user's attributes based on an input sentence. When acquiring attribute information from the platform 3, the processor 41 may acquire the attribute information from account information of the user registered on the platform 3, for example.

[0088] Next, in S240, the processor 41 executes an insertion process for inserting the appeal information of the selected appeal target product into the input text, and proceeds to S220. Examples of the appeal information include a description of the appeal target product, an image, a link to a website, and an advertisement creative. The processor 41 can execute the same process as S135 as the insertion process.

[0089] In this way, the processor 41 of the second embodiment executes processing for an input sentence sent from the platform 3 to insert promotional information for a product corresponding to the keyword into the conversation when a topic related to the input sentence is included in the table 7 and the input sentence contains a keyword corresponding to the topic of the specified conversation.

[0090] [2-2-2. Example of conversation when a topic is suggested by the user] An example of a conversation between the user and the artificial intelligence 5 when a topic is presented by the user will be described with reference to Fig. 6. Fig. 6 shows the content of the conversation between the user and the artificial intelligence 5 displayed through the platform 3 and the processing of the intermediary device 4 in the chronological order of times T1, T2, T3, ....

[0091] At time T1, the platform 3 receives an input sentence from a user through the user terminal 2. The input sentence is displayed on the user terminal 2 through the platform 3. The input sentence is further transmitted to the intermediary device 4.

[0092] At time T2, the intermediary device 4 receives the input sentence and determines that the input sentence is not the specified conversation, and therefore instructs the artificial intelligence 5 to guide the conversation to the topic "home cooking" randomly selected from table 7 (S200, S205, S210: NO in FIG. 5). The artificial intelligence 5 receives from the intermediary device 4 the instruction "Please guide the conversation to the topic of 'home cooking'. Please guide the conversation in the form of a question that can be answered with 'yes' or 'no'" along with the input sentence, and generates the sentence "Are you going to cook dinner today?" and sends it to the intermediary device 4.

[0093] At time T3, the intermediary device 4 acquires the generated sentence from the artificial intelligence 5 and transmits the generated sentence to the platform 3 (S220 in FIG. 5). At time T 4 , the generated sentence transmitted from the intermediary device 4 is displayed on the user terminal 2 via the platform 3 .

[0094] At time T5, a new input sentence from the user is input to the platform 3 through the user terminal 2. The input sentence is displayed on the user terminal 2. The input sentence is transmitted to the intermediary device 4 through the platform 3. This input sentence is expected to be a reply from the user to the generated sentence displayed at time T4.

[0095] At time T6, the intermediary device 4 receives the input sentence and determines that the input sentence is a specified conversation with the topic of "home cooking" included in table 7. Therefore, the intermediary device 4 instructs the artificial intelligence 5 to extract the keyword "recipe" or "how to make" corresponding to the topic of the input sentence "home cooking" from the conversation (S200, S205, S210: YES, S225: NO, S230 in FIG. 5). The artificial intelligence 5 receives the instruction from the intermediary device 4 along with the input sentence as a standard phrase for extracting the keyword "recipe" or "how to make", "If the conversation includes ingredients, please recommend multiple dishes using those ingredients and seasoning A. If the conversation does not include ingredients, please ask naturally about what is in the refrigerator." and generates the sentence "How about stew or kenchinjiru?" as a generated sentence and sends it to the intermediary device 4.

[0096] At time T7, the intermediary device 4, which has acquired the generated sentence from the artificial intelligence 5, transmits the generated sentence to the platform 3 (S220 in FIG. 5). At time T8, the generated sentence transmitted from the intermediary device 4 is displayed on the user terminal 2 via the platform 3.

[0097] At time T9, a new input sentence from the user is input to the platform 3 through the user terminal 2. The input sentence is displayed on the user terminal 2. The input sentence is transmitted to the intermediary device 4 through the platform 3. This input sentence is expected to be the user's reply to the generated sentence displayed at time T8.

[0098] At time T10, the intermediary device 4 receives the input sentence and determines that the input sentence is a specified conversation with the topic of "home cooking" included in table 7, and that the topic of the input sentence, "home cooking", contains the keyword "recipe", so instructs the artificial intelligence 5 to generate a sentence including the appeal information of seasoning A, which is a product corresponding to the keyword "recipe" in table 7 (S200, S205, S210: YES, S225: YES, S235 in FIG. 5). An example of the appeal information of seasoning A to be input to the artificial intelligence 5 is the name and description of seasoning A. The artificial intelligence 5 receives the instruction "Please tell me a recipe that contains seasoning A" from the intermediary device 4 together with the input sentence, and generates a sentence "I need 1 teaspoon of soy sauce, 1 teaspoon of mirin, and 1 teaspoon of seasoning A" as a generated sentence and sends it to the intermediary device 4.

[0099] At time T11, the intermediary device 4, which has acquired the generated sentence from the artificial intelligence 5, further adds appeal information for seasoning A to the generated sentence and transmits it to the platform 3 (S220 in FIG. 5). Examples of the appeal information for seasoning A that the intermediary device 4 adds to the generated sentence include a description of seasoning A, an image a of seasoning A, a link to a website related to seasoning A, "https: / / xxx.com", etc.

[0100] At time T12, the generated text to which the appeal information of seasoning A has been added and which has been sent from the intermediary device 4 is displayed on the user terminal 2 via the platform 3. In this way, when the intermediary device 4 receives an input sentence, it transfers the input sentence to the artificial intelligence 5 and outputs to the platform 3 the generated sentence generated by the artificial intelligence 5 in response to the input sentence, thereby mediating the conversation between the user and the artificial intelligence 5.

[0101] [2-2-3. Mediating when a topic is presented by artificial intelligence] The above-mentioned intermediation process is a process in which a conversation is started by transmitting an input sentence from a user. In contrast, there are cases in which a conversation is started by transmitting text from the intermediation device 4 before receiving an input sentence from a user. As an example, a conversation can be started from the intermediation device 4 at a predetermined date and time (e.g., 8:00 every morning, 17:00 every Saturday evening, etc.). The intermediation process executed by the processor 41 in this case will be described with reference to the flowchart of FIG. 7.

[0102] When starting a conversation from the intermediary device 4, the text transmitted by the intermediary device 4 is text for presenting a topic related to the product being promoted. This text may be a pre-created template, or may be text generated by the artificial intelligence 5 when the processor 41 instructs the artificial intelligence 5 to present a topic.

[0103] First, in S300, the processor 41 instructs the artificial intelligence 5 to guide the conversation to a topic selected by the processor 41 from the table 7 as a topic to be guided. When an input sentence is received from the platform 3, the processor 41 also transmits the input sentence together with the instruction to the artificial intelligence 5. The method of selecting a topic from the table 7 is not particularly limited, and for example, the processor 41 may select a topic randomly. Alternatively, each of the topics registered in the table 7 may be weighted, and the processor 41 may select a topic based on a probability corresponding to the weighting.

[0104] The method of selecting a topic based on the probability corresponding to the weighting is as follows: When a weight Wp is assigned to topic P and a weight Wq is assigned to topic Q, and the sum of the weights of all topics is ΣW, processor 41 selects topic P with a probability of Wp / ΣW and topic Q with a probability of Wq / ΣW.

[0105] When instructing the artificial intelligence 5 to guide the conversation to the guidance topic, the processor 41 may instruct the artificial intelligence 5 to perform guidance that can be answered by the user with either a positive response or a negative response. For example, the processor 41 may instruct the artificial intelligence 5 to generate a sentence that guides the conversation to the guidance topic as a question that can be answered by the user with "yes" or "no."

[0106] Next, in S305, the processor 41 receives the generated sentence from the artificial intelligence 5 and transmits it to the platform 3. Next, in S310, the processor 41 receives an input sentence from the platform 3 via the communication interface 44.

[0107] Next, in S315, the processor 41 judges whether or not the conversation has failed to be guided to the topic of the guidance target. For example, the processor 41 judges whether or not the guidance has failed depending on whether or not the topic related to the input sentence is a topic other than the topic of the guidance target for which guidance was last instructed in S300. Alternatively, the processor 41 judges whether or not the guidance has failed depending on whether or not at least one of the topics related to one or more input sentences received after the instruction to guide to the topic of the guidance target is a topic other than the topic of the guidance target. The processor 41 may input the input sentence and the topic of the guidance target for which guidance was last instructed in S300 to the artificial intelligence 5, and instruct the artificial intelligence 5 to judge whether or not the topic of the input sentence is the topic of the guidance target, and perform the judgment in S315 based on the generated sentence obtained.

[0108] When the processor 41 determines in S315 that it has failed to guide the conversation to the guidance target topic (S315: YES), the processor 41 returns to S300 and instructs the artificial intelligence 5 to guide the conversation to a topic selected by the processor 41 from the table 7 as the guidance target topic. At this time, the processor 41 selects a topic different from the topic to which the processor 41 last instructed guidance in S300.

[0109] On the other hand, when it is determined in S315 that the conversation has not failed to be guided to the guidance target topic (S315: NO), the processor 41 proceeds to S320 and determines whether or not a keyword corresponding to the guidance target topic is included in the input sentence. The keyword corresponding to the guidance target topic is a keyword associated with the guidance target topic in Table 7.

[0110] When the processor 41 determines in S320 that the keyword corresponding to the topic to be guided is not included in the input sentence (S320: NO), the processor 41 proceeds to S325. In S325, the processor 41 transmits the input sentence to the artificial intelligence 5 and instructs the artificial intelligence 5 to guide the input sentence to a topic that elicits the keyword corresponding to the topic to be guided. After that, the processor 41 returns to S305.

[0111] On the other hand, when the processor 41 determines in S320 that the keyword corresponding to the topic to be guided is included in the input sentence (S320: YES), the process proceeds to S330. In S330, the processor 41 selects a promotion target product related to the keyword from one or more promotion target products. The selection method is not particularly limited, and for example, the promotion target product may be selected randomly.

[0112] Next, in S335, the processor 41 executes an insertion process for inserting the appeal information of the selected appeal target product into the input sentence, and proceeds to S305. The processor 41 can execute the same process as S135 and S240 as the insertion process.

[0113] In this way, the processor 41 presents the user with a topic related to the product to be promoted, and starts a conversation. Then, when the processor 41 determines that the topic related to the input sentence as a reply from the user is not the presented topic (in other words, the topic to which guidance was attempted), it considers that the guidance has failed, and presents another topic to continue the conversation. On the other hand, when the processor 41 determines that the topic related to the input sentence is the presented topic, it considers that the guidance has not failed (in other words, the guidance has been successful), and guides the user to extract a keyword corresponding to that topic. When the processor 41 determines that the keyword is included in the subsequent input sentence, it executes a process for inserting promotion information of the product corresponding to that keyword into the conversation.

[0114] [2-2-4. Example of conversation when topic is presented by AI] An example of a conversation between the user and the artificial intelligence 5 when a topic is presented from the intermediary device 4 will be described with reference to Fig. 8. Fig. 8 shows the content of the conversation between the user and the artificial intelligence 5 output to the platform 3 and the processing of the intermediary device 4 in the chronological order of times T1, T2, T3, ....

[0115] At time T1, the intermediary device 4 instructs the artificial intelligence 5 to guide the conversation from the topics in the table 7 to a randomly selected topic (S300 in FIG. 7). At time T2, the intermediary device 4 acquires the generated sentence from the artificial intelligence 5 and transmits the generated sentence to the platform 3 (S305 in FIG. 7).

[0116] At time T3, the generated sentence transmitted from the intermediary device 4 is displayed on the user terminal 2 via the platform 3. At time T4, an input sentence from the user is input to platform 3 through user terminal 2. The input sentence is displayed on user terminal 2. The input sentence is transmitted to intermediary device 4 through platform 3. This input sentence is expected to be a reply from the user to the generated sentence displayed at time T3.

[0117] At time T5, the intermediary device 4 receives the input sentence and determines that the guidance has failed because the topic of the input sentence is not "exercise", which is the topic to be guided, and instructs the artificial intelligence 5 to guide the conversation to the topic "health", which is randomly selected as a topic different from the topic "exercise" in table 7 (S310, S315: YES, S300 in FIG. 7). The artificial intelligence 5 receives from the intermediary device 4 the instruction "Please guide the conversation to the topic of 'health'" along with the input sentence, and generates the sentence "How are you doing lately?" as a generated sentence and sends it to the intermediary device 4.

[0118] At time T6, the intermediary device 4, which has acquired the generated sentence from the artificial intelligence 5, transmits the generated sentence to the platform 3 (S305 in FIG. 7). At time T7, the generated sentence transmitted from the intermediary device 4 is displayed on the user terminal 2 via the platform 3.

[0119] At time T8, a new input sentence from the user is input to the platform 3 through the user terminal 2. The input sentence is displayed on the user terminal 2. The input sentence is transmitted to the intermediary device 4 through the platform 3. This input sentence is expected to be the user's reply to the generated sentence displayed at time T7.

[0120] At time T9, the intermediary device 4 receives the input sentence and determines that the guidance has not failed because the topic of the input sentence is "health," which is the topic to be induced. Furthermore, the intermediary device 4 determines that the input sentence does not contain any of the keywords "lower back pain" and "stiff shoulders," which correspond to the product medicine B corresponding to the topic "health" in table 7, and therefore instructs the artificial intelligence 5 to draw out at least one of the keywords "lower back pain" and "stiff shoulders" in the conversation (S310, S315: NO, S320: NO, S325 in FIG. 7). The artificial intelligence 5 receives from the intermediary device 4 an instruction along with the input sentence, "If you have physical pain during the conversation, naturally ask what kind of pain it is," as a standard phrase for drawing out the keyword "lower back pain" or "stiff shoulders," and generates a sentence, "What kind of pain is it?" as a generated sentence and transmits it to the intermediary device 4.

[0121] At time T10, the intermediary device 4, which has acquired the generated sentence from the artificial intelligence 5, transmits the generated sentence to the platform 3 (S305 in FIG. 7). At time T 11 , the generated sentence transmitted from the intermediary device 4 is displayed on the user terminal 2 via the platform 3 .

[0122] At time T12, a new input sentence from the user is input to the platform 3 through the user terminal 2. The input sentence is displayed on the user terminal 2. The input sentence is transmitted to the intermediary device 4 through the platform 3. This input sentence is expected to be a reply from the user to the generated sentence displayed at time T11.

[0123] At time T13, the intermediary device 4 receives the input sentence and determines that the topic of the input sentence is "health", which is the topic to be induced, and that the keyword "lower back pain" is included in the input sentence. Therefore, the intermediary device 4 selects Drug B, the product corresponding to the keyword "lower back pain" in Table 7, as the product to be promoted, and instructs the artificial intelligence 5 to generate a sentence including promotion information for Drug B (S310, S315: NO, S320: YES, S330, S335 in Figure 7).

[0124] At time T14, the intermediary device 4, which has acquired the generated sentence from the artificial intelligence 5, further adds the appeal information of the drug B to the generated sentence and transmits it to the platform 3 (S305 in FIG. 7). At time T15, the generated text to which the appeal information for drug B has been added and which has been sent from the intermediary device 4 is displayed on the user terminal 2 via the platform 3.

[0125] In this way, the intermediary device 4 selects the topic to be guided and instructs the artificial intelligence 5 to guide the conversation, and by sending the generated generated sentence to the user, the conversation is started from the intermediary device 4 without waiting for an input sentence from the user.

[0126] [2-3. Effects] According to the second embodiment described above in detail, it is possible to obtain the same effects as those of the first embodiment. In addition, according to the second embodiment, it is possible to obtain the following actions and effects.

[0127] (2a) As shown in Figures 5 and 7, the processor 41 instructs the artificial intelligence 5 to guide the conversation to a topic related to the product before inserting product appeal information into the generated sentence generated by the artificial intelligence 5. When the input sentence is a specified conversation, the processor 41 inserts the product appeal information into the generated sentence using a keyword corresponding to the topic of the specified conversation as a trigger. When the input sentence does not contain a keyword corresponding to the topic of the specified conversation, the processor 41 instructs the artificial intelligence 5 to elicit a keyword corresponding to the topic of the specified conversation from the user.

[0128] According to this configuration, it is possible to insert product appeal information that matches the topic of the conversation as an advertisement, rather than inserting product appeal information into the generated text without guiding the topic and guiding to draw out keywords. Furthermore, it is possible to insert product appeal information related to the context of the conversation, rather than inserting product appeal information into the generated text regardless of whether or not there is a keyword corresponding to the topic of the specified conversation in the conversation. Therefore, it is possible to suppress unnatural advertisement insertion in the conversation between the user and the artificial intelligence 5.

[0129] (2b) As shown in S240 of Fig. 5 and S335 of Fig. 7, when the topic of the input sentence is the topic to be induced and the input sentence contains a keyword corresponding to the topic of the designated conversation, the processor 41 instructs the artificial intelligence 5 to generate a sentence including appeal information of a product corresponding to the keyword. Furthermore, the processor 41 further adds appeal information to the generated sentence obtained from the artificial intelligence 5.

[0130] According to this configuration, appeal information that was not input to the artificial intelligence 5 can be added to the generated sentence, so that the data size of the appeal information to be input to the artificial intelligence 5 can be reduced. (2c) As shown in Figures 7 and 8, when a conversation topic is presented from the intermediary device 4, if the processor 41 determines that the topic related to the input sentence is not the last topic to which the conversation is to be guided, it instructs the artificial intelligence 5 to guide the conversation to a topic selected from the table 7 as the topic to which the conversation is to be guided. At this time, the selected topic is a topic different from the last topic to which the conversation is to be guided.

[0131] According to such a configuration, if the user is not guided to the topic to which the intermediary device 4 has guided him / her (in other words, if the guidance has failed), a different topic can be presented, thereby allowing a natural conversation with the user to continue.

[0132] (2d) The processor 41 instructs the artificial intelligence 5 to guide the conversation to a topic selected from the table 7 as a topic to be guided. The method of selecting a topic from the table 7 is such that each topic in the table 7 is weighted, and a topic is selected based on a probability corresponding to the weighting.

[0133] With this configuration, it is possible to increase the probability of directing customers to products that are more appealing. [2-4. Correspondence between terms] In the above embodiment, the input sentence corresponds to an example of a conversation, the communication interface 44 corresponds to an example of a receiving unit, the processor 41 corresponds to an example of a control unit, the table 7 corresponds to an example of topic information and first topic information, and the topic that elicits keywords corresponding to the specified conversation topic corresponds to an example of the second topic.

[0134] 3. Other embodiments Although the embodiments of the present disclosure have been described above, it goes without saying that the present disclosure is not limited to the above-described embodiments and can take various forms.

[0135] (3a) In the above embodiment, when an input sentence is received, the topic of the input sentence is identified, and it is determined whether the input sentence is a specified conversation, which is a topic included in list 6 or table 7. However, the identification of the topic is not limited to this. For example, when an input sentence is received, it may be stored as a log of the input sentence, and the topic of the conversation may be identified from the log of input sentences up to that point. Furthermore, in addition to the input sentence, the generated sentence may also be stored as a conversation log, and the topic of the conversation may be identified from the conversation log.

[0136] (3b) In the above embodiment, each time an input sentence is received, the artificial intelligence 5 is instructed to notify whether the input sentence is a specified conversation that is a topic included in the list 6 or table 7. However, the timing of issuing an instruction to the artificial intelligence 5 is not limited to each time an input sentence is received, and an instruction may be issued once each time an input sentence is received multiple times. The intermediary device 4 may issue an instruction to the artificial intelligence 5 only once at the start of a conversation.

[0137] (3c) In the above embodiment, when the intermediary device 4 determines that the input sentence is not the specified conversation, it transmits the input sentence to the artificial intelligence 5 and instructs the artificial intelligence 5 to guide the conversation to a topic selected from the list 6 as the topic to be guided. However, it is not necessary to instruct the artificial intelligence 5 to guide the conversation to the topic to be guided every time, and for example, an instruction to guide the conversation may be issued once every few times. The intermediary device 4 does not have to instruct the artificial intelligence 5 to guide the conversation to the topic to be guided. At this time, when the artificial intelligence 5 can identify that the topic of the conversation is one of the topics included in the list 6 or the table 7, it notifies the intermediary device 4 that the conversation with the user has become one of the topics included in the list 6 or the table 7.

[0138] (3d) In the above embodiment, whenever the intermediary device 4 determines that the input sentence does not contain a keyword corresponding to the specified conversation topic, it transmits the input sentence to the artificial intelligence 5 and instructs the artificial intelligence 5 to guide the conversation to a topic that elicits one or more keywords corresponding to the specified conversation topic. However, it is not necessary to instruct the artificial intelligence 5 to guide the conversation every time, and an instruction to guide the conversation may be issued once every few times. The intermediary device 4 does not have to instruct the artificial intelligence 5 to guide the conversation.

[0139] (3e) In the above embodiment, the guidance to the promoted product is divided into two stages: guidance of the topic and guidance to draw out keywords. However, the guidance to the promoted product is not limited to two stages, and may be three or more stages. Both the guidance of the topic and the guidance to the keyword may be divided into two or more stages. As an example of guidance of the topic, the topic may be guided in the order of "cooking," "cooking at home," and "menu." As an example of guidance of a keyword, in order to draw out the keyword "lower back pain," the conversation may be guided in the order of "does your body hurt?" and "where does it hurt?"

[0140] (3f) In the above embodiment, the artificial intelligence 5 identifies the topic of the conversation. However, the intermediary device 4 may identify the topic of the conversation instead of the artificial intelligence 5. For example, when the intermediary device 4 identifies a topic, it predetermines a predetermined phrase corresponding to each topic included in the list 6, and if the input sentence contains the predetermined phrase, it identifies the topic as corresponding to that phrase. If the input sentence does not contain the predetermined phrase, the topic is determined to be "unknown." For example, if the input sentence contains phrases such as "dinner" or "hungry," the topic of the input sentence can be identified as "meal," "cooking," etc.

[0141] (3g) Multiple functions possessed by one component in the above embodiments may be realized by multiple components, or one function possessed by one component may be realized by multiple components. Also, multiple functions possessed by multiple components may be realized by one component, or one function realized by multiple components may be realized by one component. Also, part of the configuration of the above embodiments may be omitted. Also, at least part of the configuration of the above embodiments may be added to or substituted for the configuration of another of the above embodiments.

[0142] [Technical idea disclosed in this specification] [Item 1] An intermediary device that mediates a conversation between a user and an artificial intelligence, A receiving unit that receives an input sentence from the user inputted via a platform; a control unit configured to transfer the input sentence received by the receiving unit to the artificial intelligence, and to output a generated sentence generated by the artificial intelligence as a response to the input sentence to the platform; Equipped with The control unit further Sending one or more topics as topic information to the artificial intelligence; Based on the topic information, on condition that the artificial intelligence has notified the user that the conversation with the user is a designated conversation that includes any one of the one or more topics, select a product to be promoted that is related to the designated conversation from among one or more products; execute an insertion process for inserting appeal information for appealing the selected appeal target product to the user into the conversation; An intermediary device configured to:

[0143] [Item 2] Item 1, an intermediary device comprising: The insertion process is a process of instructing the artificial intelligence to generate a sentence including the appeal information as a response to the input sentence. Intermediary device.

[0144] [Item 3] Item 2. An intermediary device comprising: When the control unit obtains a sentence including the appeal information from the artificial intelligence as the generated sentence, the control unit adds information including at least one of a description, an image, a link to a website, and an advertising creative related to the product to be promoted to the sentence including the appeal information, and outputs the generated sentence to the platform. Intermediary device.

[0145] [Item 4] The intermediary device according to any one of items 1 to 3, The control unit is configured to instruct the artificial intelligence to guide the conversation to any one of the one or more topics. Intermediary device.

[0146] [Item 5] Item 4. An intermediary device comprising: The prompt is a prompt to which the user can respond in either a positive or negative manner. Intermediary device.

[0147] [Item 6] The intermediary device according to item 4 or 5, The control unit selects one or more topics from the weighted topics with a probability corresponding to the weighting, and transmits the selected one or more topics to the artificial intelligence as the topic information. Intermediary device.

[0148] [Item 7] An intermediary device according to any one of items 1 to 6, The control unit selects the product to be promoted based on a keyword included in the specified conversation. Intermediary device.

[0149] [Item 8] An intermediary device according to any one of items 1 to 7, each of the one or more products having associated therewith a keyword; The control unit selects, from among the one or more products, a product associated with a keyword included in the specified conversation as the product to be promoted. Intermediary device.

[0150] [Item 9] An intermediary device according to any one of items 1 to 8, Each of the one or more topics has one or more products associated with it; The control unit, on condition that the artificial intelligence notifies the control unit that the conversation is the designated conversation, selects the product to be promoted from among the one or more products associated with the topic included in the designated conversation. Intermediary device.

[0151] [Item 10] An intermediary device according to any one of items 1 to 9, The control unit is Acquire attribute information regarding attributes of the user who converses with the artificial intelligence from at least one of the artificial intelligence and the platform; selecting the product to be promoted while taking into consideration the attributes of the user identified from the acquired attribute information; Intermediary device.

[0152] [Item 11] An intermediary device according to any one of items 1 to 10, The control unit is Sending one or more first topics to the artificial intelligence as first topic information; transmit, to the artificial intelligence, one or more second topics related to the topic included in the first designated conversation as second topic information, on condition that the artificial intelligence notifies the artificial intelligence that the conversation with the user is a first designated conversation, which is a conversation including any one of the one or more first topics, based on the first topic information; selecting, on the condition that the artificial intelligence notifies the user that the conversation with the user is a second designated conversation that includes any one of the one or more second topics based on the second topic information, from among the one or more products, a product to be promoted that is related to the second designated conversation; Intermediary device.

[0153] [Item 12] Item 12. The intermediary device according to item 11, The control unit selects the product to be promoted based on a keyword included in the second specified conversation. Intermediary device.

[0154] [Item 13] Item 11 or 12. An intermediary device according to item 11 or 12, each of the one or more products having associated therewith a keyword; The control unit selects, from among the one or more products, a product associated with a keyword included in the second specified conversation as the product to be promoted. Intermediary device.

[0155] [Item 14] An intermediary device according to any one of items 11 to 13, The control unit is Selecting a first guidance topic from among a plurality of topics, and transmitting the first guidance topic to the artificial intelligence together with information instructing the conversation to be guided to the first guidance topic; After transmitting the first guidance topic, determining whether the artificial intelligence has failed to guide the conversation to the first guidance topic; If it is determined that the conversation has failed, a topic different from the first guidance topic is selected as a second guidance topic from among the plurality of topics, and the second guidance topic is transmitted to the artificial intelligence together with information instructing the conversation to be guided to the second guidance topic. Intermediary device.

[0156] [Item 15] An intermediary device according to any one of claims 1 to 14, A storage unit configured to hold a table associating, for each of the one or more topics, one or more products related to the topic; The control unit selects the product to be promoted based on the table. Intermediary device.

[0157] [Item 16] Item 16. The intermediary device according to item 15, A plurality of topics are registered in the table, the control unit selects one or more topics from the plurality of topics registered in the table, and transmits the selected one or more topics as the topic information. Intermediary device.

[0158] [Item 17] An intermediary device that mediates a conversation between a user and an artificial intelligence, A receiving unit that receives an input sentence from the user inputted via a platform; a control unit configured to transfer the input sentence received by the receiving unit to the artificial intelligence, and to output a generated sentence generated by the artificial intelligence as a response to the input sentence to the platform; Equipped with The control unit further Sending one or more topics as topic information to the artificial intelligence; When receiving a notification from the artificial intelligence based on the topic information indicating that the conversation with the user has turned to one of the one or more topics, select a product to be promoted that is related to the topic corresponding to the notification from among one or more products; execute an insertion process for inserting appeal information for appealing the selected appeal target product to the user into the conversation; An intermediary device configured to:

[0159] [Item 18] 1. A computer-implemented method for mediating a conversation between a user and an artificial intelligence, comprising: receiving an input statement from the user inputted via a platform; Transferring the received input sentence to the artificial intelligence, and outputting a generated sentence generated by the artificial intelligence as a response to the input sentence to the platform; Sending one or more topics as topic information to the artificial intelligence; selecting, based on the topic information, from among one or more products, a product to be promoted that is related to the designated conversation, on condition that the artificial intelligence has notified the user that the conversation is a designated conversation that includes any one of the one or more topics; Executing an insertion process for inserting appeal information for appealing the selected appeal target product to the user into the conversation; , including, intermediation methods.

[0160] [Item 19] 1. A computer program for mediating a conversation between a user and an artificial intelligence, comprising: obtaining an input sentence from the user inputted via a platform; Transferring the acquired input sentence to the artificial intelligence, and outputting a generated sentence generated by the artificial intelligence as a response to the input sentence to the platform; Sending one or more topics as topic information to the artificial intelligence; selecting, based on the topic information, from among one or more products, a product to be promoted that is related to the designated conversation, on condition that the artificial intelligence has notified the user that the conversation is a designated conversation that includes any one of the one or more topics; Executing an insertion process for inserting appeal information for appealing the selected appeal target product to the user into the conversation; A computer program for causing a computer to execute the above. [Explanation of symbols]

[0161] 3...platform, 4...intermediary device, 5...artificial intelligence, 6...list, 41...processor 44...communication interface.

Claims

1. An intermediary device that mediates a conversation between a user and an artificial intelligence, A receiving unit that receives an input sentence from the user inputted via a platform; a control unit configured to transfer the input sentence received by the receiving unit to the artificial intelligence, and to output a generated sentence generated by the artificial intelligence as a response to the input sentence to the platform; Equipped with The control unit further Sending one or more topics as topic information to the artificial intelligence; Based on the topic information, on condition that the artificial intelligence has notified the user that the conversation with the user is a designated conversation that includes any one of the one or more topics, select a product to be promoted that is related to the designated conversation from among one or more products; execute an insertion process for inserting appeal information for appealing the selected appeal target product to the user into the conversation; An intermediary device configured to:

2. The intermediary device according to claim 1, The insertion process is a process of instructing the artificial intelligence to generate a sentence including the appeal information as a response to the input sentence. Intermediary device.

3. 3. The intermediary device according to claim 2, When the control unit obtains a sentence including the appeal information from the artificial intelligence as the generated sentence, the control unit adds information including at least one of a description, an image, a link to a website, and an advertising creative related to the product to be promoted to the sentence including the appeal information, and outputs the generated sentence to the platform. Intermediary device.

4. The intermediary device according to claim 1, The control unit is configured to instruct the artificial intelligence to guide the conversation to any of the one or more topics. Intermediary device.

5. The intermediary device according to claim 4, The prompt is a prompt to which the user can respond with either a positive response or a negative response. Intermediary device.

6. The intermediary device according to claim 4, The control unit selects one or more topics from the weighted topics with a probability corresponding to the weighting, and transmits the selected one or more topics to the artificial intelligence as the topic information. Intermediary device.

7. The intermediary device according to claim 1, The control unit selects the product to be promoted based on a keyword included in the specified conversation. Intermediary device.

8. The intermediary device according to claim 1, each of the one or more products is associated with a keyword; The control unit selects, from among the one or more products, a product associated with a keyword included in the specified conversation as the product to be promoted. Intermediary device.

9. The intermediary device according to claim 1, Each of the one or more topics is associated with one or more products; The control unit, on condition that the artificial intelligence notifies the control unit that the conversation is the designated conversation, selects the product to be promoted from among the one or more products associated with the topic included in the designated conversation. Intermediary device.

10. The intermediary device according to claim 1, The control unit is Acquire attribute information regarding attributes of the user who converses with the artificial intelligence from at least one of the artificial intelligence and the platform; selecting the product to be promoted while taking into consideration the attributes of the user identified from the acquired attribute information; Intermediary device.

11. The intermediary device according to claim 1, The control unit is Sending one or more first topics to the artificial intelligence as first topic information; transmit, to the artificial intelligence, one or more second topics related to the topic included in the first designated conversation as second topic information, on condition that the artificial intelligence notifies the artificial intelligence that the conversation with the user is a first designated conversation, which is a conversation including any one of the one or more first topics, based on the first topic information; selecting, on the condition that the artificial intelligence has notified the user that the conversation with the user is a second designated conversation that includes any one of the one or more second topics based on the second topic information, from among the one or more products, a product to be promoted that is related to the second designated conversation; Intermediary device.

12. The intermediary device according to claim 11, The control unit selects the product to be promoted based on a keyword included in the second specified conversation. Intermediary device.

13. The intermediary device according to claim 11, each of the one or more products is associated with a keyword; The control unit selects, from among the one or more products, a product associated with a keyword included in the second specified conversation as the product to be promoted. Intermediary device.

14. The intermediary device according to claim 1, The control unit is Selecting a first guidance topic from among a plurality of topics, and transmitting the first guidance topic to the artificial intelligence together with information instructing the conversation to be guided to the first guidance topic; After transmitting the first guidance topic, determining whether the artificial intelligence has failed to guide the conversation to the first guidance topic; If it is determined that the conversation has failed, a topic different from the first guidance target topic is selected as a second guidance target topic from among the plurality of topics, and the second guidance target topic is transmitted to the artificial intelligence together with information instructing the conversation to be guided to the second guidance target topic. Intermediary device.

15. The intermediary device according to claim 1, A storage unit configured to hold a table associating, for each of the one or more topics, one or more products related to the topic; The control unit selects the product to be promoted based on the table. Intermediary device.

16. 16. The intermediary device of claim 15, A plurality of topics are registered in the table, the control unit selects one or more topics from the plurality of topics registered in the table, and transmits the selected one or more topics as the topic information. Intermediary device.

17. An intermediary device that mediates a conversation between a user and an artificial intelligence, A receiving unit that receives an input sentence from the user inputted via a platform; a control unit configured to transfer the input sentence received by the receiving unit to the artificial intelligence, and to output a generated sentence generated by the artificial intelligence as a response to the input sentence to the platform; Equipped with The control unit further Sending one or more topics as topic information to the artificial intelligence; When receiving a notification from the artificial intelligence based on the topic information indicating that the conversation with the user has turned to one of the one or more topics, select a product to be promoted that is related to the topic corresponding to the notification from among one or more products; execute an insertion process for inserting appeal information for appealing the selected appeal target product to the user into the conversation; An intermediary device configured to:

18. 1. A computer-implemented method for mediating a conversation between a user and an artificial intelligence, comprising: receiving an input statement from the user inputted via a platform; Transferring the received input sentence to the artificial intelligence, and outputting a generated sentence generated by the artificial intelligence as a response to the input sentence to the platform; Sending one or more topics as topic information to the artificial intelligence; On the condition that the artificial intelligence notifies the user that the conversation with the user is a designated conversation including one or more topics based on the topic information, selecting a product to be promoted that is related to the designated conversation from among one or more products; Executing an insertion process for inserting appeal information for appealing the selected appeal target product to the user into the conversation; , including, intermediation methods.

19. 1. A computer program for mediating a conversation between a user and an artificial intelligence, comprising: obtaining an input sentence from the user inputted via a platform; Transferring the acquired input sentence to the artificial intelligence, and outputting a generated sentence generated by the artificial intelligence as a response to the input sentence to the platform; Sending one or more topics as topic information to the artificial intelligence; On the condition that the artificial intelligence notifies the user that the conversation with the user is a designated conversation including one or more topics based on the topic information, selecting a product to be promoted that is related to the designated conversation from among one or more products; Executing an insertion process for inserting appeal information for appealing the selected appeal target product to the user into the conversation; A computer program for causing a computer to execute the above.

Images