Automated response devices and programs

Abstract

To shorten the time required to generate a response to an inquiry. [Solution] A determination is made as to whether the inquiry content matches any of the preset inquiry contents, and if the inquiry content matches any of the preset inquiry contents, a first answer content representing a preset answer content corresponding to the preset inquiry content that matches the inquiry content is output, and if the inquiry content does not match any of the preset inquiry contents, a second answer content based on an answer content generated by a specific agent using a generation AI for the inquiry content is output.

Description

[Technical Field]

[0001] The present invention relates to a technology for performing an automatic response using generative artificial intelligence (AI). [Background technology]

[0002] There is known a technology for automatically responding using a generation AI (see, for example, Non-Patent Document 1). In such a technology, a generation AI is generally used to generate a response to a query. There is also known a Retrieval Augmented Generation (RAG) technology in which a function called an agent searches for specific data (knowledge information) based on the user's query, and generates a response using at least the search results and a generation AI. [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] "NTT Technocross Utilizes NTT's Proprietary LLM "tsuzumi" for "ForeSight Voice Mining"," [online], January 31, 2024, Rick Telecom Co., Ltd., CALL CENTER JAPAN, [Retrieved August 14, 2024], Internet<https: / / callcenter-japan.com / article / 6643 / 1 / > Summary of the Invention [Problem to be solved by the invention]

[0004] By using generation AI to generate answers to inquiries, it is possible to provide appropriate responses to a variety of inquiries. However, generating answers using generation AI can sometimes take time.

[0005] The present invention provides a technique for shortening the time required to obtain a response to an inquiry. [Means for solving the problem]

[0006] In the present invention, it is determined whether the inquiry content matches any of the preset inquiry contents, and if the inquiry content matches any of the preset inquiry contents, a first answer content representing a preset answer content corresponding to the preset inquiry content that matches the inquiry content is output, and if the inquiry content does not match any of the preset inquiry contents, a second answer content based on an answer content generated by a specific agent using a generation AI for the inquiry content is output. [Effects of the Invention]

[0007] This reduces the time required to obtain a response to an inquiry. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 is a block diagram illustrating the configuration of an automatic answering system according to an embodiment. [Figure 2] FIG. 2 is a block diagram illustrating a terminal device according to the embodiment. [Figure 3] FIG. 3 is a block diagram illustrating the automatic answering device of the embodiment. [Figure 4] FIG. 4 is a diagram illustrating examples of preset inquiry contents and corresponding preset response contents. [Figure 5] FIG. 5 is a flow diagram illustrating the automatic response process according to the embodiment. [Figure 6] FIG. 6 is a diagram illustrating a display screen according to the embodiment. [Figure 7] FIG. 7 is a diagram illustrating a display screen according to the embodiment. [Figure 8] FIG. 8 is a block diagram illustrating a hardware configuration of the embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0009] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. [First embodiment] <Overall structure> As illustrated in FIG. 1, the automatic answering system 1 of this embodiment includes terminal devices 11-1, ..., 11-M, an automatic answering device 12, and a generation AI device 13. Hereinafter, each of the terminal devices 11-1, ..., 11-M will be referred to as terminal device 11-m. Here, m = 1, ..., M, and M is an integer greater than or equal to 1, for example, an integer greater than or equal to 2. Furthermore, the terminal devices 11-1, ..., 11-M will be collectively referred to as terminal device 11. Each terminal device 11-m is a device used by user 1011-m. Hereinafter, users 1011-1, ..., 1011-M will be collectively referred to as user 1011. The automatic answering device 12 is configured to be able to communicate with the terminal device 11 via a network such as the Internet. Furthermore, the automatic answering device 12 is configured to be able to call the generation AI device 13 and use the generation AI device 13. User 1011-m uses terminal device 11-m to make an inquiry to automatic answering device 12, and automatic answering device 12 obtains an answer to the inquiry using generation AI device 13 and responds to terminal device 11-m. Note that the automatic answering system 1 illustrated in Figure 1 has one automatic answering device 12 and one generation AI device 13. However, automatic answering system 1 may have multiple automatic answering devices 12 and multiple generation AI devices 13.

[0010] <Terminal Device 11-m> As illustrated in FIG. 2, the terminal device 11-m of this embodiment includes a user interface 111-m, a communication unit 112-m, a control unit 116-m, and a memory 117-m. The terminal device 11-m executes various processes under the control of the control unit 116-m. Information input to the terminal device 11-m and information obtained by each unit of the terminal device 11-m are stored in the memory 117-m and are read out and used as needed. Examples of the terminal device 11-m include smartphones, smartwatches, tablet terminals, and personal computers loaded with a predetermined program. The user interface 111-m is, for example, a touch screen or a combination of a display, a mouse, a keyboard, a touch panel, a speaker, a microphone, and the like.

[0011] <Automatic Answering Machine 12> As illustrated in FIG. 3, the automatic answering device 12 of this embodiment includes a storage unit 120, a communication unit 121, a calling unit 123, a dispatcher 124, agents 125-1, ..., 125-N, a control unit 126, and a memory 127. N is an integer equal to or greater than 1, for example, N is an integer equal to or greater than 2. Each of the agents 125-1, ..., 125-N is referred to as agent 125-n, where n = 1, ..., N. The agents 125-1, ..., 125-N are collectively referred to as agent 125. A set having agents 125-1, ..., 125-N as elements is referred to as an agent set. The dispatcher 124 includes a determination unit 1241, response units 1242 and 1244, and a selection unit 1243. The agent 125-n includes a search unit 1251-n and a generation unit 1252-n. The search units 1251-1, ..., 1251-N are collectively referred to as search unit 1251, and the generation units 1252-1, ..., 1252-N are collectively referred to as generation unit 1252. The automatic answering device 12 executes each process under the control of the control unit 126. Information input to the automatic answering device 12 and information obtained by each unit of the automatic answering device 12 are stored in memory 127 and are read out and used as needed. Examples of the automatic answering device 12 include a server device into which a predetermined program has been loaded and a device configured on the cloud.

[0012] <Generation AI device 13> The generative AI device 13 is a device that implements a generative AI based on a machine learning model and has the function of generating and returning a response to given information. For example, the generative AI generates text based on a large-scale language model (LLM). In addition, the generative AI may generate images based on an image generation model, generate videos based on a video generation model, generate audio based on an audio generation model, or generate music based on a music generation model. Examples of the generative AI device 13 include a server device into which a predetermined program is loaded, a device configured on the cloud, etc.

[0013] <Pre-settings> The following pre-settings are prerequisites for the automatic response process of this embodiment. (1) Pre-settings for Agent 125 (2) Pre-settings for Dispatcher 124

[0014] (1) Pre-settings for Agent 125 The agent 125 is a processing unit that searches preset data (knowledge information) based on the query content, obtains search results, and obtains a response corresponding to the query content using at least the search results and the generation AI. For example, the agent 125 may provide the search results or information based thereon to the generation AI and obtain a response content from the generation AI, or may provide the query content and the search results or information based thereon to the generation AI and obtain a response content from the generation AI. Data searchable by each agent 125-n is preset, which determines the properties of each agent 125-n. In other words, the data searchable by each agent 125-n functions as the knowledge of that agent 125-n, and the response content obtained using at least the search results and the generation AI is based on the knowledge of that agent 125-n. In other words, setting the data searchable by each agent 125-n is equivalent to setting the knowledge of each agent 125-n. Preferably, data searchable by the same agent 125 has commonalities, while data searchable by different agents 125 has differences. For example, the same agent 125 may be able to search for data in the same field, while different agents 125 may be able to search for data in different fields. Specifically, for example, agent 125-1 may be able to search for only data in the IT field, agent 125-2 may be able to search for only data in the legal field, and agent 125-3 may be able to search for only data in the economics field. For example, the same agent 125 may be able to search for data related to the same product, while different agents 125 may be able to search for data related to different products. Specifically, for example, agent 125-1 may be able to search for only data related to telephone and fax multifunction devices, agent 125-2 may be able to search for only data related to smartphones, and agent 125-3 may be able to search for only data related to automobiles. For example, the same agent 125 may be able to search for data related to the same model, while different agents 125 may be able to search for data related to different models.For example, agent 125-1 may be able to search only for data related to model A of a telephone and fax multifunction device, agent 125-2 may be able to search only for data related to model B of a telephone and fax multifunction device, and agent 125-3 may be able to search only for data related to model C of a telephone and fax multifunction device. The data may be document data, text data, tabular data, spreadsheet data, or table data divisible by columns and rows, text data divisible by chapters, paragraphs, sentences, etc., tagged XML (Extensible Markup Language) data, comma separated values ​​(CSV) file data divisible by entries, program code data divisible by comments, graphic data, video data, or audio data. The data may be private or public information. In this embodiment, at least one of the data includes FAQ data representing a pair of a preset inquiry and a preset answer corresponding to the preset inquiry (hereinafter referred to as a "pair of a preset inquiry and an answer"). A set of preset inquiries and answers is called, for example, a "Frequently Asked Questions (FAQ)." Figure 4 shows examples of sets of preset inquiries and answers. Figure 4 shows examples of a set of preset inquiries, such as "What kind of service is 'XXX'?" and a set of preset answers, such as "XXX is a value plan that bundles six optional services and call charges." Also shown are a set of a preset inquiry, such as "What are the conditions for free use of 'XXX'?" and a set of preset answers, such as "To get free use of 'XXX', you must sign up for 'XXX'." In addition, data may represent, for example, the business knowledge of a specific organization, know-how, specialized knowledge, or a product manual.

[0015] In the presetting of the agent 125, data searchable by the agent 125-n (n = 1, ..., N) is stored in the memory 120 of the automatic answering device 12. Hereinafter, data searchable by the agent 125-n will be referred to as data 1212-n, and data 1212-1, ..., 1212-N will be collectively referred to as data 1212. At least a portion of the data 1212-1, ..., 1212-N stored in the memory 120 includes FAQ data. In other words, the data 1212 searchable by at least one of the one or more agents 125 includes information representing a preset inquiry content and information representing a preset answer content. In addition to the data 1212-n, other registration information 1213-n related to the data 1212-n and / or the agent 125-n may be stored in the memory 120. For example, the registration information 1213-n may include summary information corresponding to data searchable by the agent 125-n, or may include information representing search conditions (search parameters) when searching for data. The summary information may be, for example, summary information of the agent 125-n, or summary information of the data 1212-n searchable by the agent 125. For example, the summary information is information searched to select an agent suitable for the content of an inquiry.

[0016] (2) Pre-settings for Dispatcher 124 The dispatcher 124 in this embodiment is a processing unit that provides general information for automatic responses. The dispatcher 124 has a function of using input information according to the inquiry content to search for information corresponding to the data 1212 searchable by the agent 125, thereby selecting a specific agent 125 from which to obtain a response to the inquiry content (generated response content). Furthermore, the dispatcher 124 has a function of directly accessing at least FAQ data among the data 1212 stored in the storage unit 120. When the input inquiry content matches any FAQ data, the dispatcher 124 obtains the response content represented by the FAQ data matching the input inquiry content, instead of having the agent 125 obtain the response content. This reduces the time required to obtain a response to the inquiry content. In the pre-settings of the dispatcher 124, criteria for the dispatcher 124 to search for information are set. For example, criteria for the dispatcher 124 to search for FAQ data may be set, or criteria for the dispatcher 124 to search for an agent 125 may be set. The criteria for the search may include, for example, the allowable error (tolerance) in the vector search, or may include conditions for the text search, such as an exact match search or a partial match search.

[0017] <Automatic response processing> Next, the automatic response process of this embodiment will be described with reference to FIG. The user 1011-m inputs the inquiry content into the user interface 111-m of the terminal device 11-m (FIG. 2). The inquiry content may be input, for example, by text input or voice input. For example, a display screen 1100 illustrated in FIGS. 6 and 7 is displayed on the user interface 111-m, and the user 1011-m inputs the inquiry content into an input field 1110 on the display screen 1100. For example, in the example of FIG. 6, the inquiry content "What is . . . ." is input, and in the example of FIG. 7, the inquiry content "Regarding . . . ., what happens if . . . and . . . are added?" is input. Information representing the input inquiry content is sent to the communication unit 112-m. The communication unit 112-m transmits the information representing the inquiry content to the automatic answering device 12 via a network. The communication unit 121 of the automatic answering device 12 (FIG. 3) receives the information representing the inquiry content and sends it to the dispatcher 124. The determination unit 1241 of the dispatcher 124 receives information indicating the content of the inquiry (step S101).

[0018] The determination unit 1241 of the dispatcher 124 accesses the FAQ data included in the data 1212 stored in the storage unit 120, and determines whether or not the inquiry content sent from the communication unit 121 matches any of the preset inquiry content (inquiry content of the FAQ data). For example, the determination unit 1241 of the dispatcher 124 determines whether or not the inquiry content matches any of the preset inquiry content based on the determination criteria preset as described above (step S102).

[0019] If it is determined in step S102 that the inquiry matches any of the preset inquiry contents, the determination unit 1241 extracts, from the storage unit 120, preset answer contents corresponding to the preset inquiry contents that match the inquiry content, and sends the answer contents (first answer contents) representing the extracted answer contents to the answer unit 1242 (first answer unit). The answer unit 1242 outputs the answer contents (first answer contents) representing the extracted answer contents to the user 1011-m to the communication unit 121. This answer contents to the user 1011-m may include additional information indicating that it is a preset answer contents. The communication unit 121 transmits information representing the answer contents to the terminal device 11-m via the network. The communication unit 112-m of the terminal device 11-m (FIG. 2) receives the information representing the answer contents to the user 1011-m and sends it to the user interface 111-m. The user interface 111-m outputs the answer contents to the user 1011-m. For example, the user interface 111-m may display the answer to the user 1011-m or output it as audio. For example, in the example of FIG. 6, the display screen 1100 is displayed on the user interface 111-m, and the answer (first answer) is displayed in the answer field 1120. In this example, the answer displayed in the answer field 1120 is, "'XXX' is a value plan that includes six optional services and call charges." Furthermore, if the answer to the user 1011-m includes additional information indicating that the answer is a preset answer, the user interface 111-m may output, in addition to the answer, a message indicating that the answer is a preset answer. In the example of FIG. 6, a mark 1121 "FAQ" indicating that the answer is a preset answer is displayed in the answer field 1120. Then, the process proceeds to step S107 (step S103).

[0020] On the other hand, if it is determined in step S102 that the query content does not match any of the preset query content, the selection unit 1243 of the dispatcher 124 selects a specific agent 125-s suited to the query content from among the agents 125-1, ..., 125-N (s∈{1, ..., N}). The specific agent 125-s belongs to a set of one or more agents 125-1, ..., 125-N. If N=1, the selection unit 1243 selects the agent 125-1. If N≧2, the selection unit 1243 selects a specific agent 125-s suited to the query content from among the multiple agents 125-1, ..., 125-N based on input information corresponding to the query content. That is, the selection unit 1243 selects the specific agent 125-s by using the input information corresponding to the query content to search for information corresponding to the data 1212 that each of the multiple agents can search. This search may be, for example, a vector search, a keyword search, another search, or a hybrid search that combines multiple search methods. The specific agent 125-s to be selected may be singular or plural. The number C of specific agents 125-s may be set, or an upper limit Cmax of the number of specific agents 125-s may be set, or these may not be set. Note that Cmax is an integer equal to or greater than 1. Below, an example of a method for selecting a specific agent 125-s when N≧2 is shown.

[0021] <Example 1 of how to select a specific Agent 125-s> (1-1) The selection unit 1243 uses a search query representing the inquiry content (input information according to the inquiry content) to search for information corresponding to the data 1212-1, ..., 1212-N stored in the storage unit 120. The information corresponding to the data 1212-1, ..., 1212-N may be, for example, the data 1212-1, ..., 1212-N itself, function values ​​(excluding constant function values) derived from each of the data 1212-1, ..., 1212-N, the registered information 1213-1, ..., 1213-N, a portion (e.g., summary information) of each of the registered information 1213-1, ..., 1213-N, function values ​​(excluding constant function values) derived from each of the registered information 1213-1, ..., 1213-N, or function values ​​(excluding constant function values) derived from a portion (e.g., summary information) of each of the registered information 1213-1, ..., 1213-N. The function value may be a scalar value or a vector value. For example, the selection unit 1243 uses a vector representing the inquiry content as a search query and performs a vector search for vectors representing summary information of each of the registered information 1213-1, ..., 1213-N. The criteria for the search may be fixed or may be preset as described above. (1-2) If the number of hits in step (1-1) exceeds the upper limit Cmax, proceed to the following step (1-3). On the other hand, if the number of hits in step (1-1) is equal to or less than the upper limit Cmax, proceed to the following step (1-5). (1-3) The selection unit 1243 communicates (for example, communicates via chat) with the terminal device 11-m to narrow down the number of hits. That is, the selection unit 1243 sends information representing questions for narrowing down the number of hits to the communication unit 121, and the communication unit 121 transmits the information representing the questions to the terminal device 11-m via the network. The communication unit 112-m of the terminal device 11-m (FIG. 2) receives the information representing the questions and transmits it to the user interface 111-m, which then outputs the questions. The user 111-m inputs answers to the questions into the user interface 111-m. The answers are transmitted to the communication unit 112-m and transmitted to the automatic answering device 12 via the network. The communication unit 121 of the automatic answering device 12 (FIG. 3) receives the answers and transmits them to the selection unit 1243. The selection unit 1243 uses a search query representing information corresponding to the response content and the inquiry content (input information corresponding to the inquiry content) to search for information corresponding to the data 1212-1, ..., 1212-N stored in the storage unit 120. This search method is the same as in step (1-1). (1-4) If the number of hits in step (1-3) exceeds the upper limit Cmax, return to step (1-3). On the other hand, if the number of hits in step (1-3) is equal to or less than the upper limit Cmax, proceed to the following step (1-5). (1-5) The selection unit 1243 selects the agent 125 corresponding to the information found in the search as a specific agent 125-s.

[0022] <Example 2 of how to select a specific Agent 125-s> (2-1) The selection unit 1243 uses a search query representing the inquiry content (input information corresponding to the inquiry content) to search for information corresponding to the data 1212-1, ..., 1212-N stored in the storage unit 120. This search method is the same as in step (1-1). (2-2) The selection unit 1243 selects Cmax or less agents 125 from among the agents 125 corresponding to the information hit in the search in order of the degree of matching (for example, relevance score, similarity, etc.) and outputs them as specific agents 125-s. For example, when the selection unit 1243 performs a vector search for vectors (for example, vectors representing summary information) corresponding to the data 1212-1, ..., 1212-N using a vector representing the inquiry content as a search query, the selection unit 1243 selects Cmax or less agents 125 corresponding to vectors selected in order of the degree of similarity to the search query as specific agents 125-s.

[0023] <Example 3 of how to select a specific Agent 125-s> Information representing the specific agent 125-s selected in Example 1 or Example 2 of the selection method for the specific agent 125-s may be sent to the terminal device 11-m, and the agents 125-s may be narrowed down based on the selection by the user 1011-m. That is, the selection unit 1243 sends the information representing the agent 125-s to the communication unit 121, and the communication unit 121 transmits the information representing the agent 125-s to the terminal device 11-m via the network. The communication unit 112-m of the terminal device 11-m (FIG. 2) receives the information representing the agent 125-s and transmits it to the user interface 111-m, which then outputs the information representing the agent 125-s. The user 1011-m selects some or all of the output agents 125-s and inputs information representing the selection into the user interface 111-m. The information representing the selection is sent to the communication unit 112-m and transmitted to the automatic answering device 12 via the network. The communication unit 121 of the automatic answering device 12 (FIG. 3) receives the information representing the selected content and sends it to the selection unit 1243. The selection unit 1243 sets the final specific agent 125-s based on the information representing the selected content.

[0024] The selection unit 1243 (FIG. 3) sends information indicating the above-mentioned specific agent 125-s to the reply unit 1244 (step S104).

[0025] The answering unit 1244 (second answering unit) outputs an answer (second answering content) based on the generated answer content acquired by a specific agent 125-s (s∈{1,...,N}) using the generation AI for the query content. That is, each of the agents 125 has a function of acquiring the generated answer content using at least the search results obtained by searching the searchable data 1212 based on the query content and the generation AI. The answering unit 1244 controls the specific agent 125-s to acquire the generated answer content for the query content and output the answer content based on the generated answer content. For example, the answering unit 1244 sends information representing the query content to each of the specific agents 125-s described above and makes an inquiry to each agent 125-s. The search unit 1251-s of the agent 125-s searches the data 1212-s stored in the storage unit 120 using the information representing the query content and extracts search results, which are hit information. This search may be, for example, a vector search, a keyword search, or other search, or a hybrid search that combines multiple search methods. The search criteria may be fixed or may be expressed in the registration information 1213-s. It is desirable that the search results be divisible into semantically coherent units based on their structure. Examples of such search results include tabular data, spreadsheet data, and table data (columns and rows), text data (chapters, paragraphs, sentences, etc.), XML data (tags), CSV file data (entries), and program code data (commented code). The search results may include a single semantically coherent unit or multiple semantically coherent units. The search results are sent to the generation unit 1252-s of the agent 125-s. The generation unit 1252-s calls the generation AI device 13 via the calling unit 123 (API (Application Programming Interface)) and obtains a generated response content corresponding to the query content using information representing the query content, data based on the search results, and the generation AI of the generation AI device 13.That is, the generation unit 1252-s obtains a generated response content corresponding to the query content using at least the search results and the generation AI. For example, the generation unit 1252-s sends data (e.g., payload) based on information representing the query content and the search results to the generation AI device 13 via the call unit 123. The generation AI device 13 generates a generated response content for the data based on the information representing the query content and the search results, and returns the generated response content (e.g., payload) to the generation unit 1252-s via the call unit 123. Note that if the search results can be divided into semantically coherent units by structure, the generation unit 1252-s preferably obtains a generated response content corresponding to the query content using at least the search results (e.g., rows of tabular data, etc.) divided into semantically coherent units by structure and the generation AI. The generation unit 1252-s sends the obtained generated response content to the response unit 1244 of the dispatcher 124. The response unit 1244 of the dispatcher 124 obtains and outputs a response content (second response content) for the user 1011-m based on the generated response content sent from each agent 125-s. For example, the response unit 1244 may use a summary of the generated response content sent from the agent 125-s as the response content for the user 1011-m, or may use any of the generated response content sent from the agent 125-s as the response content for the user 1011-m, or may use new response content generated from the generated response content sent from the agent 125-s as the response content for the user 1011-m. For example, the response unit 1244 may call the generation AI device 13 via the calling unit 123 and obtain a response content for the user 1011-m using the generated response content sent from the agent 125-s and the generated AI of the generation AI device 13. This response content for the user 1011-m may include additional information indicating that it is based on content generated by the generation AI. Information representing the content of the answer to the user 1011-m is sent to the communication unit 121. The communication unit 121 transmits the information representing the content of the answer to the user 1011-m to the terminal device 11-m via the network. The communication unit 112-m of the terminal device 11-m (FIG. 2) receives the information representing the content of the answer to the user 1011-m and sends it to the user interface 111-m.The user interface 111-m outputs the answer content to the user 1011-m. For example, the user interface 111-m may display the answer content to the user 1011-m or may output it as audio. For example, the display screen 1100 illustrated in FIG. 7 is displayed on the user interface 111-m, and the answer content (second answer content) is displayed in the answer field 1120 of the display screen 1100. In the example of FIG. 7, the answer content "Regarding , if , and are added ..." is displayed in the answer field 1120. Furthermore, if the answer content to the user 1011-m includes additional information indicating that it is based on content generated by the generation AI, the user interface 111-m may output, in addition to the answer content, a message indicating that the answer content is based on content generated by the generation AI. In the example of FIG. 7, an "AI" mark 1122 indicating that the answer content is based on content generated by the generation AI is displayed in the answer field 1120. Then, the process proceeds to step S107 (step S105).

[0026] In step S107, the determination unit 1241 (FIG. 3) determines whether a new inquiry has been received from the user 1011-m. For example, after the answering unit 1242 or the answering unit 1244 outputs an answer to the user 1011-m (step S103 or S105), the determination unit 1241 determines that a new inquiry has been received from the user 1011-m when the determination unit 1241 receives information indicating a new inquiry from the terminal device 11-m. For example, when the display screen 1100 illustrated in FIG. 6 or FIG. 7 is displayed on the user interface 111-m of the terminal device 11-m (FIG. 2) (step S103 or S105) and the answer is displayed in the answer field 1120, the user 1011-m can input a new inquiry into the input field 1130 using the user interface 111-m. In the examples of FIGS. 6 and 7, the user 1011-m may input the content of the inquiry into the input field 1130, may select the content of the inquiry from preset inquiry contents (e.g., frequently asked questions) displayed in the tag section 1131, or may add the inputted content to the inquiry content selected from the tag section 1131. The preset inquiry content displayed in the tag section 1131 may or may not be the same as that represented by the FAQ data included in the data 1212 (FIG. 3). Information representing the inputted content of the inquiry is sent to the communication unit 112-m of the terminal device 11-m (FIG. 2). The communication unit 112-m transmits the information representing the inquiry to the automatic response unit 12 via the network. The communication unit 121 of the automatic response unit 12 (FIG. 3) receives the information representing the inquiry and sends it to the dispatcher 124. The determination unit 1241 of the dispatcher 124 accepts the information representing the inquiry. For example, in this case, the determining unit 1241 determines that a new inquiry has been received from the user 1011-m. If it is determined that a new inquiry has been received from the user 1011-m, the process returns to step S102.On the other hand, for example, if the determination unit 1241 does not receive information indicating new inquiry content from the terminal device 11-m even after a predetermined time has elapsed after the reply content to the user 1011-m is output from the reply unit 1242 or 1244 (step S103 or S105), the determination unit 1241 determines that new inquiry content has not been received from the user 1011-m. In this case, the automatic response process ends.

[0027] <Features of this form> In this embodiment, the determination unit 1241 of the dispatcher 124 of the automatic answering device 12 (FIG. 3) determines whether the inquiry content matches any of the "preset inquiry content." If the inquiry content matches any of the "preset inquiry content," the reply unit 1242 (first reply unit) outputs reply content (first reply content) representing the "preset reply content" corresponding to the "preset inquiry content" that matches the inquiry content. This reduces the time required to acquire the reply content for the inquiry content. On the other hand, if the inquiry content does not match any of the "preset inquiry content," the reply unit 1244 (second reply unit) outputs reply content (second reply content) based on the generated reply content acquired by a specific agent 125-s using a generation AI for the inquiry content. This allows appropriate responses to a variety of inquiries. In this way, this embodiment allows appropriate responses to a variety of inquiries and reduces the time required to acquire the reply content for the inquiry content.

[0028] In this embodiment, a specific agent 125-s belongs to a set of one or more agents 125-1, ..., 125-N, and each of the one or more agents 125 has a function of acquiring a generated answer content using at least a search result obtained by searching searchable data 1212 based on the inquiry content and a generation AI, and the data 1212 (knowledge information) searchable by at least one of the one or more agents 125 includes information representing preset inquiry content and information representing preset answer content (FAQ data). That is, the answer content (second answer content) based on the generated answer content acquired by the agent is based on the data 1212 (knowledge information), and the answer content (first answer content) representing the preset answer content is based on the FAQ data included in the data 1212 (knowledge information). Therefore, consistency is improved between the answer content (first answer content) representing the preset answer content and the answer content (second answer content) based on the generated answer content acquired by the agent, improving the overall sense of unity of the answer content.

[0029] [Second embodiment] The second embodiment is a modification of the first embodiment. The FAQ data included in the data 1212 varies in completeness. The amount of answer content in the FAQ data included in each data 1212-n may be large or small, and the quality of the answer content may be high or low. Generally, the more complete the answer content of the FAQ data, the more suitable it is as an answer to the user 1011. Taking this into consideration, the dispatcher of this embodiment determines whether the inquiry content matches any of the preset inquiry content using a determination criterion based on an index representing the completeness of the preset answer content (FAQ data). Here, the frequency with which the preset answer content with the index having a first value is determined to match the inquiry content is higher than the frequency with which the preset answer content with the index having a second value is determined to match the inquiry content. The completeness represented by the first value is higher than the completeness represented by the second value. This improves the balance between the quality of the answer content and the short response time, improving the overall quality of the automatic response. The following description will focus on the differences from the first embodiment, and the same reference numbers will be used to simplify the description of items already described.

[0030] <Overall structure> As illustrated in FIG. 1, the automatic response system 2 of this embodiment includes terminal devices 11-1, ..., 11-M, an automatic response device 22, and a generation AI device 13. The automatic response device 22 is configured to be able to communicate with the terminal device 11 via a network such as the Internet. The automatic response device 22 is also configured to be able to call the generation AI device 13 and use the generation AI device 13. A user 1011-m makes an inquiry to the automatic response device 22 using the terminal device 11-m, and the automatic response device 22 obtains a response to the inquiry using the generation AI device 13 and responds to the terminal device 11-m. Note that the automatic response system 2 illustrated in FIG. 1 includes one automatic response device 22 and one generation AI device 13. However, the automatic response system 2 may include multiple automatic response devices 22 and multiple generation AI devices 13.

[0031] <Automatic Answering Machine 22> 3, the automatic answering device 22 of this embodiment has a storage unit 120, a communication unit 121, a calling unit 123, a dispatcher 224, agents 125-1, ..., 125-N, a control unit 126, and a memory 127. The dispatcher 224 has a determination unit 2241, response units 1242 and 1244, and a selection unit 1243. The automatic answering device 22 executes each process under the control of the control unit 126. Information input to the automatic answering device 22 and information obtained by each unit of the automatic answering device 22 are stored in the memory 127 and are read out and used as needed. Examples of the automatic answering device 22 include a server device into which a predetermined program is loaded, a device configured on the cloud, etc.

[0032] <Pre-settings> The differences from the presetting in the first embodiment are as follows.

[0033] (1) Pre-settings for Agent 125 In this embodiment, at least the registered information 1213-n is stored in the storage unit 120. However, in this embodiment, each registered information 1213-n includes at least an index i(n) representing the degree of completeness of the preset answer content included in each data 1212-n. The index i(n) representing the degree of completeness may represent, for example, the quantity or quality of the preset answer content included in the data 1212-n, or a combination thereof. For example, the greater the quantity of the answer content, the higher the degree of completeness, or the higher the quality of the answer content, the higher the degree of completeness. The quantity of the preset answer content may represent, for example, the number of preset answer content (e.g., the number of pairs of preset inquiry content and answer content), the quantity (e.g., the total number of characters in the answer content), the number of link destinations included in the answer content (e.g., the total number of links included in the answer content), or at least a combination thereof. The quality of the preset answer content may represent, for example, the satisfaction level of the answer content, the rate at which the inquiry is completed by the answer content, or a combination of these. The rest is the same as the preset of the agent 125 in the first embodiment, except that the automatic answering device 12 is replaced with the automatic answering device 22.

[0034] (2) Pre-settings for Dispatcher 224 The dispatcher 224 of this embodiment determines whether the inquiry matches any of the preset inquiry contents based on a determination criterion based on an index i(n) that indicates the completeness of the preset answer content. As a result of this determination, the frequency with which the preset answer content (FAQ data included in the data 1212-n) for which the index i(n) is α1 (first value) is determined to match the inquiry content is higher than the frequency with which the preset answer content (FAQ data included in the data 1212-n) for which the index i(n) is α2 (second value) is determined to match the inquiry content. Note that the completeness represented by α1 (first value) is higher than the completeness represented by α2 (second value). For example, the more complete the FAQ data is, the more frequently it is determined to match the inquiry content. At least this determination criterion is set in the pre-setting of the dispatcher 224 of this embodiment. The rest is the same as the pre-setting of the dispatcher 124 of the first embodiment, except that the automatic response device 12 is replaced with the automatic response device 22.

[0035] Below, examples of judgment criteria based on the index i(n) representing the degree of fulfillment are given. <Examples of criteria> For example, the index i(n) representing the degree of fulfillment is determined as follows: i(n) = (update frequency score × W1) + (average feedback score × W2) + (number of items score × W3) In this example, the larger the index i(n), the greater the degree of completeness of the FAQ data contained in the data 1212-n. The update frequency score is, for example, a score that indicates the update frequency of the FAQ data included in the data 1212-n. The higher the update frequency score, the higher the update frequency of the FAQ data. For example, the update frequency score is determined as follows: Recently updated: +10 points A few months ago: +5 points Over a year ago: +0 points When the data 1212-n includes a plurality of FAQ data, for example, the total value or average value of the update frequency scores of the plurality of FAQ data may be used. The average feedback score is, for example, the average of the feedback scores for the answers represented by the FAQ data included in the data 1212-n. The higher the feedback score, the better the evaluation of the answers. For example, the average feedback score is determined as follows: ·4.5-5.0: +10 points ·3.5-4.4: +8 points ·2.5-3.4: +5 points ·1.5-2.4: +2 points ·1.0-1.4: +0 points The number score is, for example, a score based on the number of FAQ data items included in the data 1212-n. The higher the FAQ number score, the greater the number of FAQ data items included in the data 1212-n. For example, the number score is determined as follows: ·30 or more: +10 points 10 to 30 results: +5 points Less than 10: +0 points W1, W2, and W3 are weights that represent the importance of each score. For example, the weights W1, W2, and W3 are determined as follows: W1 (Importance of update frequency score): 0.3 W2 (mean feedback score importance): 0.6 W3 (Importance of number score): 0.1 For example, if the update frequency score is +10 points, the average feedback score is +8 points, the number of items score is +10 points, and W1 = 0.3, W2 = 0.6, and W3 = 0.1, the index i(n) representing the degree of fulfillment is calculated as follows: i(n)=(10×0.3)+(8×0.6)+(10×0.1)=3+4.8+1=8.8 The dispatcher 224 of this embodiment determines whether the inquiry matches any of the preset inquiry contents, for example, using a determination criterion based on such index i(n). For example, if the similarity between the inquiry and the preset inquiry exceeds a threshold based on index i(n), the dispatcher 224 determines that the inquiry matches the preset inquiry content, and otherwise determines that the inquiry does not match. However, the threshold is set to be smaller as the index i(n) increases. As a result, the frequency with which a preset answer content with index i(n) α1 is determined to match the inquiry content is higher than the frequency with which a preset answer content with index i(n) α2 is determined to match the inquiry content. The degree of fulfillment represented by α1 is higher than the degree of fulfillment represented by α2.

[0036] <Automatic response processing> The automatic response process of this embodiment is the same as the first embodiment, except that step S102 of the automatic response process of the first embodiment is replaced with the following step S202 (FIG. 5). The automatic response device 12 and the dispatcher 124 are replaced with the automatic response device 22 and the dispatcher 224, respectively.

[0037] Step S202: The determination unit 2241 of the dispatcher 224 (FIG. 3) accesses the registration information 1213 and the FAQ data included in the data 1212 stored in the storage unit 120, and determines whether the inquiry content sent from the communication unit 121 matches any of the preset inquiry contents (inquiry contents of the FAQ data) using a determination criterion based on the index i(n) representing the degree of completeness represented by the registration information 1213-n. This determination results in a higher frequency of determining that a preset answer content whose index i(n) is α1 (first value) matches the inquiry content than a higher frequency of determining that a preset answer content whose index i(n) is α2 (second value) matches the inquiry content. Note that the degree of completeness represented by α1 (first value) is higher than the degree of completeness represented by α2 (second value). The determination criterion based on the index i(n) representing the degree of completeness is as described above. For example, in the case of the example of the judgment criteria described above, if the similarity between the query content and the preset query content exceeds a threshold based on the index i(n), the judgment unit 2241 of the dispatcher 224 judges that the query content matches the preset query content, and otherwise judges that the query content does not match (step S202).

[0038] <Features of this form> In this embodiment, the determination unit 2241 of the dispatcher 224 of the automatic answering device 22 (FIG. 3) determines whether the inquiry content sent from the communication unit 121 matches any of the "preset inquiry contents" using a determination criterion based on an index i(n) representing the degree of completeness of the "preset answer contents." In this embodiment, the frequency with which a preset answer content with an index i(n) of α1 (first value) is determined to match the inquiry content is higher than the frequency with which a preset answer content with an index i(n) of α2 (second value) is determined to match the inquiry content. Note that the degree of completeness represented by α1 (first value) is higher than the degree of completeness represented by α2 (second value). By using the determination criterion based on the degree of completeness of the "preset answer contents" in this manner, it is possible to appropriately determine whether to output an answer content representing the "preset answer contents" (first answer content) or an answer content based on the generated answer content acquired by a specific agent 125-s (second answer content). As a result, the balance between the quality of the response and the short response time is improved, and the quality of the automatic response is improved overall.

[0039] [Third embodiment] For the same user 1011, before an answer based on the generated answer is output to a certain inquiry, an answer (third answer) representing a preset answer to a past inquiry (second inquiry) may be output (FIG. 5). In such a case, if the latter answer (third answer) output earlier can be used to acquire the former generated answer, the consistency of the answer in a series of automatic response processes can be improved. From this perspective, the agent of this embodiment acquires the generated answer to the former inquiry by using at least the latter answer (third answer), the search results, and the generation AI. Note that the latter answer (third answer) represents a preset answer (answer content of the FAQ data) corresponding to a preset inquiry that matches an earlier inquiry (second inquiry) than the former inquiry. The following description will focus on differences from the first embodiment, and the same reference numbers will be used to simplify the description of matters already described.

[0040] <Overall structure> As illustrated in FIG. 1, the automatic response system 3 of this embodiment includes terminal devices 11-1, ..., 11-M, an automatic response device 32, and a generation AI device 13. The automatic response device 32 is configured to be able to communicate with the terminal device 11 via a network such as the Internet. The automatic response device 32 is also configured to be able to call the generation AI device 13 and use the generation AI device 13. A user 1011-m makes an inquiry to the automatic response device 32 using the terminal device 11-m, and the automatic response device 32 obtains a response to the inquiry using the generation AI device 13 and responds to the terminal device 11-m. Note that the automatic response system 3 illustrated in FIG. 1 includes one automatic response device 32 and one generation AI device 13. However, the automatic response system 3 may include multiple automatic response devices 32 and multiple generation AI devices 13.

[0041] <Automatic Answering Machine 32> 3, the automatic answering device 32 of this embodiment has a storage unit 120, a communication unit 121, a calling unit 123, a dispatcher 324, agents 125-1, ..., 125-N, a control unit 126, and a memory 127. The dispatcher 324 has a determination unit 1241, response units 3242, 3244, and a selection unit 1243. The automatic answering device 32 executes each process under the control of the control unit 126. Information input to the automatic answering device 32 and information obtained by each unit of the automatic answering device 32 are stored in the memory 127 and are read out and used as needed. Examples of the automatic answering device 32 include a server device into which a predetermined program is loaded, a device configured on the cloud, etc.

[0042] <Pre-settings> The presetting is the same as that of the first embodiment, except that the automatic answering device 12 is replaced with an automatic answering device 32.

[0043] <Automatic response processing> The automatic response process of this embodiment is the same as the first embodiment, except that step S103 of the automatic response process of the first embodiment is replaced with the following step S303, and step S105 is replaced with the following step S305. The automatic response process of this embodiment is the same as the first embodiment, except that the automatic response unit 12 and the dispatcher 124 are replaced with the automatic response unit 32 and the dispatcher 324, respectively.

[0044] Step S303: In step S303, instead of the answering unit 1242, the answering unit 3242 (first answering unit) executes the processing of step S103 described in the first embodiment, and further stores the answer content (third answer content) for the user 1011-m acquired in the processing of step S103 in the memory 127 (step S303).

[0045] Step S305: In step S305, instead of the answering unit 1244, the answering unit 3244 (second answering unit) outputs an answer (second answering content) based on the generated answer content acquired by a specific agent 125-s (s∈{1, ..., N}) using the generation AI for the query content. However, each of the agents 125 in this embodiment has a function of acquiring the generated answer content using at least past answer content (third answer content) extracted from the memory 127 in addition to the search results obtained by searching the searchable data 1212 based on the query content and the generation AI. Note that the query content corresponding to the generated answer content and the past query content (second query content) corresponding to the past answer content (third answer content) used to acquire the generated answer content are preferably transmitted from the same terminal device 11. In other words, the query content corresponding to the acquired generated answer content and the past query content (second query content) are preferably generated by the same user 1011. More preferably, the inquiry content corresponding to the generated answer content to be acquired and the past inquiry content (second inquiry content) are inquiry content received by the dispatcher 324 from the time the dispatcher 324 receives the inquiry content in step S101 until it is determined in step S107 that the dispatcher 324 has not received information representing new inquiry content. In other words, these inquiry contents are preferably those received by the dispatcher 324 in a series of automatic response processes. However, this does not limit the present invention. For example, the inquiry content corresponding to the generated answer content to be acquired and the past inquiry content (second inquiry content) may be generated by a user 1011 with the same attribute (i.e., they may be sent from a terminal device 11 of a user 1011 with the same attribute). Examples of attributes include nationality, occupation, field of expertise, age, gender, and length of use period. For example, the answering unit 3244 sends the past answer content (third answer content) and information representing the inquiry content to each of the above-mentioned specific agents 125-s and makes an inquiry to each agent 125-s.The search unit 1251-s of the agent 125-s uses information representing the query content to search the data 1212-s stored in the memory unit 120 and extracts search results, which are hit information. This search is, for example, as described in the first embodiment. The search results are sent to the generation unit 1252-s of the agent 125-s. The generation unit 1252-s calls the generation AI device 13 via the calling unit 123 and obtains a generated answer content corresponding to the query content using the information representing the query content, data based on past answer content (third answer content), search results, and the generated AI of the generation AI device 13. That is, the generation unit 1252-s obtains a generated answer content corresponding to the query content using at least the past answer content (third answer content), search results, and the generation AI. For example, the generation unit 1252-s sends data based on the information representing the query content, past answer content (third answer content), and search results to the generation AI device 13 via the calling unit 123. The generation AI device 13 generates a generated answer content for the data based on the information representing the inquiry content, the past answer content (third answer content), and the search results, and returns it to the generation unit 1252-s via the calling unit 123. The generation unit 1252-s sends the obtained generated answer content to the answer unit 3244 of the dispatcher 324. The processing of step S305 thereafter is the same as the processing of step S105 described in the first embodiment, except that the dispatcher 124 is replaced by the dispatcher 324 and the answer unit 1244 is replaced by the answer unit 3244 (step S305).

[0046] <Features of this form> A specific agent 125-s (s∈{1,...,N}) of this embodiment acquires a generated answer content using at least an answer content (third answer content) representing a "preset answer content" corresponding to a "preset inquiry content" that matches an inquiry content (second inquiry content) that is older than the current inquiry content, and a generating AI. The generated answer content thus obtained reflects the past answer content (third answer content), thereby improving the consistency of the answer content in a series of automatic response processes.

[0047] [Modification 1 of the third embodiment] In the third embodiment, if it is determined in step S102 (FIG. 5) that the inquiry does not match any of the preset inquiry contents, the selection unit 1243 of the dispatcher 324 selects a specific agent 125-s that is suitable for the inquiry content from among the agents 125-1, ..., 125-N (s∈{1, ..., N}). Here, the selection unit 1243 may select the specific agent 125-s based on the above-mentioned past answer content (a third answer content representing a preset answer content corresponding to a preset inquiry content that matches a second inquiry content that is older than the inquiry content). For example, the selection unit 1243 may select an agent 125 that can search for FAQ data corresponding to the past answer content (third answer content) as the specific agent 125-s. Alternatively, for example, the selection unit 1243 may select, as the specific agent 125-s, an agent 125 selected by the method exemplified in the first embodiment (examples 1 to 3 of the method for selecting the specific agent 125-s) and an agent 125 capable of searching for FAQ data corresponding to the past answer content (third answer content). Alternatively, for example, when selecting an agent 125 by the method exemplified in the first embodiment (examples 1 to 3 of the method for selecting the specific agent 125-s), the selection unit 1243 may use search criteria that make it easier to select an agent 125 capable of searching for FAQ data corresponding to the past answer content (third answer content) (step S304'). This improves the consistency between the past answer content (third answer content) and the answer content newly generated using the generation AI.

[0048] [Modification 2 of the third embodiment] The dispatcher 324 of the third embodiment includes the determination unit 1241, the response units 3242 and 3244, and the selection unit 1243. However, the determination unit 1241 of the dispatcher 324 may be replaced with the determination unit 2241 of the second embodiment. In this case, the presetting described in the second embodiment is executed, and step S102 is replaced with step S202 described in the second embodiment (FIG. 5). Furthermore, step S304' described in the first modification of the third embodiment may be executed instead of step S104.

[0049] [Modification 3 of the third embodiment] In the third embodiment and its modified examples 1 and 2, the past answer content (third answer content) refers to the answer content of the FAQ data (preset answer content corresponding to the preset inquiry content that matches the past second inquiry content). However, in the third embodiment and its modified examples 1 and 2, the past answer content (third answer content) may be the answer content based on the generated answer content acquired by a specific agent 125-s in response to the past inquiry content (steps S105, S305).

[0050] [Hardware configuration] The functions performed by the components described herein may be implemented in circuitry or processing circuitry, including general-purpose processors, application-specific processors, integrated circuits, ASICs (Application Specific Integrated Circuits), a CPU (a Central Processing Unit), conventional circuits, and / or combinations thereof, programmed to perform the described functions. A processor includes transistors and other circuits and is considered to be circuitry or processing circuitry. A processor may also be a programmed processor that executes programs stored in memory.

[0051] In this specification, a circuitry, unit, or means is hardware that is programmed to realize or performs the described functions, which may be any hardware disclosed herein or any hardware known to be programmed to realize or perform the described functions.

[0052] If the hardware is a processor considered to be a type of circuitry, the circuitry, means, or unit is a combination of the hardware and software used to configure the hardware and / or processor.

[0053] For example, the terminal device 11 and the automatic answering devices 12, 22, and 32 in each embodiment are devices configured by a general-purpose or dedicated computer having a processor (hardware processor) such as a CPU (central processing unit) and memories such as RAM (random-access memory) and ROM (read-only memory) executing a predetermined program. That is, the terminal device 11 and the automatic answering devices 12, 22, and 32 in each embodiment have, for example, processing circuitry configured to implement each of the components possessed by the terminal device 11 and the automatic answering devices 12, 22, and 32. This computer may have one processor and memory, or multiple processors and memories. This program may be installed on the computer or may be pre-recorded in a ROM or the like. Furthermore, some or all of the processing units may be configured using electronic circuits that independently realize processing functions, rather than electronic circuits that realize functional configurations by loading programs, such as a CPU. Furthermore, the electronic circuits constituting one device may include multiple CPUs.

[0054] FIG. 8 is a block diagram illustrating the hardware configuration of the terminal device 11 and the automatic answering devices 12, 22, and 32 in each embodiment. As illustrated in FIG. 8, the terminal device 11 and the automatic answering devices 12, 22, and 32 in this example include a central processing unit (CPU) 10a, an input unit 10b, an output unit 10c, a random access memory (RAM) 10d, a read-only memory (ROM) 10e, an auxiliary storage device 10f, a communication unit 10h, and a bus 10g. The CPU 10a in this example includes a control unit 10aa, a calculation unit 10ab, and a register 10ac, and executes various calculation processes according to various programs loaded into the register 10ac. The input unit 10b is an input terminal, keyboard, mouse, touch panel, etc., through which data is input. The output unit 10c is an output terminal, display, etc., through which data is output. The communication unit 10h is a LAN card, etc., controlled by the CPU 10a that has loaded a predetermined program. The RAM 10d is a static random access memory (SRAM), a dynamic random access memory (DRAM), or the like, and has a program area 10da where a predetermined program is stored and a data area 10db where various data are stored. The auxiliary storage device 10f is a hard disk, a magneto-optical disc (MO), a semiconductor memory, or the like, and has a program area 10fa where a predetermined program is stored and a data area 10fb where various data are stored. The bus 10g connects the CPU 10a, the input unit 10b, the output unit 10c, the RAM 10d, the ROM 10e, the communication unit 10h, and the auxiliary storage device 10f so that information can be exchanged. The CPU 10a writes the program stored in the program area 10fa of the auxiliary storage device 10f to the program area 10da of the RAM 10d in accordance with the loaded OS (Operating System) program. Similarly, the CPU 10a writes various data stored in the data area 10fb of the auxiliary storage device 10f to the data area 10db of the RAM 10d.The addresses in RAM 10d where the programs and data are written are stored in register 10ac of CPU 10a. Control unit 10aa of CPU 10a sequentially reads out these addresses stored in register 10ac, reads out the programs and data from the areas in RAM 10d indicated by the read addresses, causes calculation unit 10ab to sequentially execute the calculations indicated by the programs, and stores the calculation results in register 10ac. This configuration realizes the functional configuration of terminal device 11 and automatic answering devices 12, 22, and 32.

[0055] The program describing this processing can be recorded on a computer-readable recording medium. Examples of computer-readable recording media are non-transitory recording media. Examples of such recording media include magnetic recording devices, optical disks, magneto-optical recording media, and semiconductor memories.

[0056] The program may be distributed, for example, by selling, transferring, lending, etc. a portable recording medium such as a DVD or CD-ROM on which the program is recorded. Furthermore, the program may be stored in a storage device of a server computer, and then transferred from the server computer to another computer via a network, thereby distributing the program.

[0057] A computer that executes such a program may first temporarily store the program recorded on a portable recording medium or transferred from a server computer in its own storage device. Then, when executing a process, the computer reads the program stored on its own recording medium and executes the process in accordance with the read program. Alternatively, the computer may read the program directly from a portable recording medium and execute the process in accordance with the program. Furthermore, the computer may execute the process in accordance with the program each time a program is transferred from a server computer to the computer. The server computer may not transfer the program to the computer, but may instead execute the process through a so-called ASP (Application Service Provider) service, which realizes the processing function by issuing an execution instruction and obtaining the results. Furthermore, the server computer may execute the process on a terminal using a so-called SaaS (Software as a Service) service, which allows users to use part of the server computer along with the program. In this embodiment, the program includes information used for computer processing that is equivalent to a program (such as data that is not a direct instruction to the computer but has properties that define computer processing).

[0058] Furthermore, in this embodiment, the device is configured by executing a predetermined program on a computer, but at least a part of the processing contents may be realized by hardware.

[0059] [Other variations] It should be noted that the present invention is not limited to the above-described embodiment. For example, in the above-described embodiment, the determination units 1241, 2241 (FIG. 3) directly access the FAQ data included in the data 1212 stored in the storage unit 120 and determine whether the inquiry matches any of the preset inquiry contents. However, the FAQ data included in the data 1212 may be duplicated in the pre-setting, and the duplicated FAQ data may be stored in the storage unit 120 or another storage unit. In this case, the determination units 1241, 2241 may access the duplicated FAQ data rather than directly accessing the FAQ data included in the data 1212, and determine whether the inquiry matches any of the preset inquiry contents.

[0060] Furthermore, the various processes described above may not only be executed in chronological order as described, but may also be executed in parallel or individually depending on the processing capabilities of the devices executing the processes or as necessary. Needless to say, other modifications are possible within the scope of the present invention.

[0061] [Note] The above contents can be summarized as follows: [Appendix 1] a determination unit that determines whether the inquiry content matches any of the preset inquiry contents; a first reply unit that, when the inquiry content matches any of the preset inquiry contents, outputs a first reply content that represents a preset reply content corresponding to the preset inquiry content that matches the inquiry content; and a second answering unit that outputs a second answer based on a generated answer obtained by a specific agent using a generation AI for the inquiry content when the inquiry content does not match any of the pre-set inquiry contents. [Appendix 2] 1. The automated answering device of claim 1, the particular agent belongs to a set of one or more agents; each of the one or more agents has a function of acquiring a generated answer content using at least a search result obtained by searching searchable data based on the inquiry content and the generating AI; The data searchable by at least one of the one or more agents includes information representing the preset inquiry content and information representing the preset answer content. [Appendix 3] An automated answering device according to claim 1 or 2, the determination unit determines whether the inquiry matches any of the preset inquiry contents based on a determination criterion that is based on an index that indicates the completeness of the preset answer content; the frequency with which the preset answer content is determined to match the inquiry content when the index has a first value is higher than the frequency with which the preset answer content is determined to match the inquiry content when the index has a second value; The degree of fulfillment represented by the first value is higher than the degree of fulfillment represented by the second value. [Appendix 4] An automatic answering device according to any one of appendices 1 to 3, The specific agent obtains the generated answer content using at least a third answer content representing a preset answer content corresponding to a preset inquiry content that matches a second inquiry content that is older than the inquiry content, and the generation AI. [Appendix 5] An automatic answering device according to any one of appendices 1 to 4, The automatic answering device further includes a selection unit that selects the specific agent based on a third response content that represents a preset response content corresponding to a preset inquiry content that matches a second inquiry content that is older than the inquiry content. [Appendix 6] An automatic answering device according to any one of appendices 1 to 5, The automatic answering device includes additional information indicating that the first response content is a preset response content. [Appendix 7] 6. An automated answering device according to claim 4 or 5, The automatic answering device, wherein the inquiry content and the second inquiry content are transmitted from the same terminal device. [Appendix 8] 6. An automated answering device according to claim 4 or 5, The automatic answering device, wherein the inquiry content and the second inquiry content are transmitted from terminal devices of users having the same attribute. [Appendix 9] An automatic answering method executed by an automatic answering device, comprising: a determination step in which a determination unit determines whether or not the inquiry content matches any of the preset inquiry contents; a first reply step in which, when the inquiry content matches any of the preset inquiry contents, a first reply unit outputs a first reply content that represents a preset reply content corresponding to the preset inquiry content that matches the inquiry content; a second answering step in which, when the inquiry content does not match any of the preset inquiry contents, a second answering unit outputs a second answer content based on a generated answer content obtained by a specific agent using a generation AI for the inquiry content; An auto-answer method having [Appendix 10] A program for causing a computer to function as an automatic answering device as set forth in any of appendices 1 to 9. [Explanation of symbols]

[0062] 12,22,32 Auto Answering Machine 124,224,324 Dispatcher 1241,2241 Judgment section 1242,3242 Answer part (1st answer part) 1244,3244 Answer part (2nd answer part) 1243 Selection Section

Claims

1. a determination unit that determines whether the inquiry content matches any of the preset inquiry contents; a first reply unit that, when the inquiry content matches any of the preset inquiry contents, outputs a first reply content that represents a preset reply content corresponding to the preset inquiry content that matches the inquiry content; a second answering unit that outputs a second answer based on a generated answer obtained by a specific agent using a generation AI for the inquiry content when the inquiry content does not match any of the preset inquiry contents; the particular agent belongs to a set of one or more agents; each of the one or more agents has a function of acquiring a generated response content by using at least a search result obtained by searching searchable data based on the inquiry content and the generation AI; The data searchable by at least one of the one or more agents includes information representing the preset inquiry content and information representing the preset answer content.

2. a determination unit that determines whether the inquiry content matches any of the preset inquiry contents; a first reply unit that, when the inquiry content matches any of the preset inquiry contents, outputs a first reply content that represents a preset reply content corresponding to the preset inquiry content that matches the inquiry content; a second answering unit that outputs a second answer based on a generated answer obtained by a specific agent using a generation AI for the inquiry content when the inquiry content does not match any of the preset inquiry contents; the determination unit determines whether the inquiry matches any of the preset inquiry contents based on a determination criterion that is based on an index that indicates the completeness of the preset answer content; the frequency with which the preset answer content when the index has a first value is determined to match the inquiry content is higher than the frequency with which the preset answer content when the index has a second value is determined to match the inquiry content, The degree of fulfillment represented by the first value is higher than the degree of fulfillment represented by the second value.

3. a determination unit that determines whether the inquiry content matches any of the preset inquiry contents; a first reply unit that, when the inquiry content matches any of the preset inquiry contents, outputs a first reply content that represents a preset reply content corresponding to the preset inquiry content that matches the inquiry content; a second answering unit that outputs a second answer based on a generated answer obtained by a specific agent using a generation AI for the inquiry content when the inquiry content does not match any of the preset inquiry contents; The specific agent obtains the generated answer content using at least a third answer content representing a preset answer content corresponding to a preset inquiry content that matches a second inquiry content that is older than the inquiry content, and the generation AI.

4. a determination unit that determines whether the inquiry content matches any of the preset inquiry contents; a first reply unit that, when the inquiry content matches any of the preset inquiry contents, outputs a first reply content that represents a preset reply content corresponding to the preset inquiry content that matches the inquiry content; a second answering unit that outputs a second answer based on a generated answer obtained by a specific agent using a generation AI for the inquiry content when the inquiry content does not match any of the preset inquiry contents; The automatic answering device further includes a selection unit that selects the specific agent based on a third response content that represents a preset response content corresponding to a preset inquiry content that matches a second inquiry content that is older than the inquiry content.

5. A program for causing a computer to function as the automatic answering device according to any one of claims 1 to 4.

Images