Information processing apparatus, information processing method, and information processing program

Abstract

To improve retrieval performance for creating an appropriate answer to a question of a user.SOLUTION: An information processing apparatus according to the present application includes a retrieval unit and a specification unit. The retrieval unit retrieves a second generation target that is a generation target having high degree of similarity to information based on a question of a user and is a generation target generated based on a first generation target. The specification unit specifies the first generation target serving as a basis for creating an answer to the question based on the second generation target retrieved by the retrieval unit.SELECTED DRAWING: Figure 10

Description

[Technical Field]

[0001] The present invention relates to an information processing device, an information processing method, and an information processing program. [Background technology]

[0002] Chatbots and other automated response services have become increasingly popular. For example, when a user inputs a question, a search model is used to generate an answer by referencing sentences corresponding to the user's question. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2023-076413 Summary of the Invention [Problem to be solved by the invention]

[0004] However, conventional technologies sometimes fail to provide an appropriate answer to a user's question, for example, when the user's question is abstract, when there are many similar sentences referenced when creating an answer to a question, or when the source of the answer is unknown and it is difficult to determine whether it is correct or incorrect.

[0005] The present application has been made in view of the above, and aims to improve search performance for generating appropriate answers to user questions. [Means for solving the problem]

[0006] The information processing device of the present application is characterized by having a search unit that searches for second-generation objects that are generation objects that have a high similarity to information based on a user's question and are generated based on first-generation objects, and an identification unit that identifies the first-generation objects that serve as a basis for creating an answer to the question based on the second-generation objects searched for by the search unit. [Effects of the Invention]

[0007] According to one aspect of the embodiment, it is possible to improve search performance for creating an appropriate answer to a user's question. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 is an explanatory diagram for explaining an overview of information processing according to an embodiment. [Figure 2] FIG. 2 is a diagram illustrating an example of the configuration of an information processing system according to the embodiment. [Figure 3] FIG. 3 is a diagram illustrating an example of information processing according to the embodiment. [Figure 4] FIG. 4 is a diagram illustrating an example of information processing (question conversion) according to the embodiment. [Figure 5] FIG. 5 is a diagram illustrating an example of information processing (search refinement) according to the embodiment. [Figure 6A] FIG. 6A is an explanatory diagram (1) for explaining the relationship between a first-generation target and a second-generation target according to the embodiment. [Figure 6B] FIG. 6B is an explanatory diagram (2) for explaining the relationship between the first generation target and the second generation target according to the embodiment. [Figure 7] FIG. 7 is a diagram illustrating an example of information processing (reranking) according to the embodiment. [Figure 8] FIG. 8 is an explanatory diagram for explaining reranking according to the embodiment. [Figure 9] FIG. 9 is a diagram illustrating an example of the configuration of a terminal device according to the embodiment. [Figure 10] FIG. 10 is a diagram illustrating an example of the configuration of an information processing device according to the embodiment. [Figure 11] FIG. 11 is a diagram illustrating an example of a first-generation target storage unit according to the embodiment. [Figure 12] FIG. 12 is a diagram illustrating an example of a second-generation target storage unit according to the embodiment. [Figure 13]FIG. 13 is a flowchart illustrating an example of the question conversion process according to the embodiment. [Figure 14] FIG. 14 is a flowchart illustrating an example of a search refinement process according to the embodiment. [Figure 15] FIG. 15 is a flowchart illustrating an example of the reranking process according to the embodiment. [Figure 16] FIG. 16 is a hardware configuration diagram illustrating an example of a computer that realizes the functions of the information processing device. DETAILED DESCRIPTION OF THE INVENTION

[0009] Hereinafter, an information processing device, an information processing method, and an information processing program according to the present application (hereinafter referred to as "embodiments") will be described in detail with reference to the drawings. Note that the information processing device, the information processing method, and the information processing program according to the present application are not limited to these embodiments. Furthermore, the same components in the following embodiments will be denoted by the same reference numerals, and duplicated descriptions will be omitted.

[0010] (Embodiment) [1. Overview of Information Processing] Conventionally, automated response services such as chatbots have become increasingly popular. For example, a technology is known in which, when a user inputs a question, a search model is used to generate an answer by referencing sentences corresponding to the user's question. However, conventional technologies may not be able to provide an appropriate answer to the user's question in some cases, such as when the user's question is abstract, when there are many similar sentences referenced when generating an answer to the question, or when the source of the answer is unknown and it is difficult to determine whether it is correct or incorrect.

[0011] For example, if a user asks a question such as "What are the company's employee benefits?", the system may not be able to provide an answer if the question is too abstract and there are no search hits, or if no related sentences can be found. Also, if there are many similar sentences and it is difficult to determine which one to use, the system may use all of them, regardless of whether they are related to the answer. Furthermore, if the source of the answer is unknown, the system may use sentences that are not directly mentioned, such as translations, to provide an answer.

[0012] Fig. 1 shows an example of information processing according to the prior art. When a user U1 inputs a question Q about a company's internal regulations (step S1), multiple sentences highly similar to the question Q are searched for and extracted from the internal regulations (step S2) (step S3), and an answer to the question Q is generated from the sentences and provided (step S4).

[0013] However, as described above, the information processing shown in FIG. 1 may not be able to provide an appropriate answer to a user's question. In particular, so-called generative AI, such as a sentence generation model such as GPT or Stable Diffusion, is trained using information belonging to many domains to generate and output responses (text and images) to various inputs. As a result of such training, generative AI can output general answers to general questions, but cannot generate appropriate answers to questions that require information in a specific domain, such as company regulations. On the other hand, a model that has trained only on information in a specific domain has not learned general responses, and therefore requires the user to provide appropriate input. Therefore, a process is required that can appropriately provide specialized information, such as company regulations, as an answer to a question entered by a user without considering the input required to obtain an appropriate answer. The present application has been made in view of the above, and aims to improve search performance for generating appropriate answers to user questions.

[0014] [2. Information Processing System Configuration] An information processing system 1 shown in Fig. 2 will be described. As shown in Fig. 2, the information processing system 1 includes a terminal device 10 and an information processing device 100. The terminal device 10 and the information processing device 100 are connected to each other via a predetermined communication network (network N) so as to be able to communicate with each other via wired or wireless communication. Fig. 2 is a diagram showing an example of the configuration of the information processing system 1 according to an embodiment.

[0015] The terminal device 10 is an information processing device used by a user who asks a question. The user may ask a question about the contents of regulations, laws, rules, manuals, etc. For example, a user may ask a question such as "Please tell me about the employee benefits offered at the company" to check the employee benefits offered at the company. The user may input a question using, for example, a predetermined chat tool. The terminal device 10 may be any device that can implement the processing in the embodiment. The terminal device 10 may also be a device such as a smartphone, a tablet device, a notebook PC, a desktop PC, a mobile phone, or a PDA. FIG. 2 shows a case where the terminal device 10 is a smartphone.

[0016] The terminal device 10 is, for example, a smart device such as a smartphone or tablet, and is a mobile terminal device that can communicate with any server device via a wireless communication network such as 4G to 5G (Generations) or LTE (Long Term Evolution). The terminal device 10 may have a screen such as a liquid crystal display with a touch panel function, and may accept various operations on displayed data such as content, such as tapping, sliding, and scrolling, performed by a user using a finger or a stylus. In FIG. 2, the terminal device 10 is used by a user U1.

[0017] The information processing device 100 is an information processing device intended to improve search performance for creating appropriate answers to user questions, and may be any device capable of implementing the processes described in the embodiments. For example, the information processing device 100 aims to provide appropriate answers to user questions even when the user's question is abstract, when there are many similar sentences referenced when creating an answer to a question, or when the source of the answer is unknown and it is difficult to determine whether it is correct or incorrect. The information processing device 100 performs at least one of the following three main processes (question conversion, search refinement, and reranking). For example, the information processing device 100 may perform at least one of the following processes, while another information processing device may perform the other processes, and the processing results may be provided to the information processing device 100.

[0018] In the first process (question conversion), the information processing device 100 provides the user's question to the generation AI to make the question more specific. This is believed to enable appropriate search hits even when the question is abstract, and also to enable appropriate retrieval of related sentences. In the second process (search refinement), the information processing device 100 uses the result of the first process to search for information highly similar to the first process result and identify information that can serve as the basis for creating an answer to the question. This is believed to enable an answer using sentences that are truly relevant to the answer to be obtained even when there are many similar sentences. In the third process (reranking), the information processing device 100 uses the result of the second process to evaluate the relevance of the second process result to the question and rerank the second process result. This is believed to enable an answer using an appropriate source to be obtained even when the source is unknown and it is difficult to determine whether it is correct or not. Then, the information processing device 100 provides the result of the third process to the generation AI to generate and provide an answer to the question.

[0019] Generative AI is a generative model trained on data published on the Internet, for example. One example is a GPT (Generative Pre-trained Transformer) model that generates answers to questions. Such models generate answers to questions from data published on the Internet. Since new data is published on the Internet every day, the answers provided by the model are updated daily. In other words, even if the same question is asked, different answers will be generated depending on the time of the question, as the data referenced when the model generates the answer will differ. Furthermore, since the model can provide detailed answers to questions in a wide range of fields, users ask the model a wide variety of questions.

[0020] Although FIG. 2 shows a case where the terminal device 10 and the information processing device 100 are separate devices, the terminal device 10 and the information processing device 100 may be integrated.

[0021] [3. An example of information processing] FIG. 3 is a diagram illustrating an example of information processing according to an embodiment. The diagram summarizes the three processes described above and shows a flow of a series of processes. Details of each of the three processes will be described later. When a user U1 inputs a question Q about a company's internal regulations (step S11), the question Q is provided to a generation AI (step S12) to convert the question (to specify the question) (step S13). Then, the AI ​​compares the similarity (cosine similarity) with the information after the question to search for multiple sentences with high similarity (step S14), and identifies sentences linked to each of the sentences (step S15). The AI ​​then provides the identified sentences to the generation AI (step S16) and reranks them by evaluating their relevance to the question Q (step S17). Based on the reranking results, sentences with high relevance ratings are extracted (step S18), and an answer to the question Q is generated from the extracted sentences and provided (step S19). Details of each process are described below.

[0022] FIG. 4 is a diagram illustrating an example of information processing (question conversion) according to an embodiment. The information processing device 100 acquires a question from the user U1, which is transmitted from the terminal device 10 when the user U1 inputs a question (step S101). In FIG. 4, a question Q is acquired. The information processing device 100 provides the question Q to the generation AI along with a prompt for specifying the question Q (step S102), and acquires a generation result from the generation AI (step S103). Note that this generation result is information for narrowing down targets that will be used as a basis for creating an answer to the question Q in the second process. That is, the information processing device 100 uses the generation result in step S103 to search for information that will be used as a basis for creating an answer to the question Q (step S104). Note that this question conversion process may be a process executed when the question has a high level of abstraction. For example, the information processing device 100 may evaluate the level of abstraction of the question Q, and perform question conversion if the level of abstraction of the question Q is evaluated as high, and perform the second process using the question Q if the level of abstraction of the question Q is not evaluated as high.

[0023] Here, details of question conversion according to the embodiment will be described. There are mainly two types of question conversion according to the embodiment. The first question conversion is conversion into a provisional answer. For example, if the question Q is "Tell me about employee benefits," the question Q is converted into a provisional answer such as "Our company places great emphasis on employee benefits. Below are our main employee benefit systems. Health insurance: Our company provides health insurance to all employees. This allows employees to cover medical expenses in the event of illness or injury. These are our main employee benefit systems. We are enhancing these systems to make it easier for employees to work and increase their satisfaction."

[0024] The second type of question transformation is to transform the question into a similar question. For example, it transforms into one or more questions. For example, if the question Q is "Tell me about employee benefits," it transforms into one or more similar questions such as "What are the types and contents of employee benefits?", "Tell me about the company's employee benefits system," "What are the benefits of employee benefits?", and "Tell me how to introduce employee benefits and give some specific examples."

[0025] The information processing device 100 may perform question conversion by providing the generation AI with a prompt for converting question Q into a provisional answer together with question Q, or by providing the generation AI with a prompt for converting question Q into a similar question together with question Q. The prompt may also include an instruction sentence indicating the instruction to convert question Q into a provisional answer, as well as an instruction sentence expressing at least one of the role, task content, task objective, premise, and constraint in natural language. Here, the role is a sentence indicating the position in which the generation AI will generate an answer, such as a sentence that sets the role, such as "You are a professional prompt generator." The task content is a sentence indicating the content of the task the generation AI will perform when generating an answer, such as a sentence that instructs the task, such as "Your task is to write a sentence that answers the user's question." The task objective is a sentence indicating the purpose of the task the generation AI will perform when generating an answer, such as a sentence that provides a task objective, such as "To help the user search for related sentences." The premise is a sentence that indicates the premise under which the generation AI performs the task when generating an answer, such as "This is related to the procedures of the company's internal regulations." The constraint is a sentence that indicates the constraint under which the generation AI performs the task when generating an answer, such as "Write within XX characters." The conversion of question Q into a tentative answer may be performed based on instructions such as the role, task content, task purpose, premise, and constraint. The prompt may include an instruction sentence indicating an instruction to convert question Q into a similar question, as well as an instruction sentence expressing at least one of role, task content, task purpose, premise, and constraint in natural language. The conversion of question Q into a similar question may be performed based on instructions such as the role, task content, task purpose, premise, and constraint.

[0026] As a result, vector search is difficult for abstract questions, but by making the question more specific, it is possible to improve search accuracy.

[0027] FIG. 5 is a diagram showing an example of information processing (search refinement) according to an embodiment. In the following embodiment, an object that is the basis for creating an answer to question Q is referred to as a first-generation object, and an object generated from the first-generation object is referred to as a second-generation object. The first-generation object and the second-generation object have a parent chunk and child chunk relationship. For example, the second-generation object is a generation object generated by dividing the first-generation object into predetermined sentence units. Also, for example, the second-generation object is a generation object generated by dividing the first-generation object into sentence units so that the number of characters is equal to or less than the maximum number. Also, the first-generation object is a generation object generated from predetermined source data on the assumption that it will be used to answer a question.

[0028] In Figure 5, to determine candidate information to be used in creating an answer to question Q, candidate information to be used in creating an answer to question Q is directly searched for from the information type "first-generation target," and information for identifying candidate information to be used in creating an answer to question Q is searched for from the information type "second-generation target." Here, as a result of searching for information from the information type "first-generation target" and the information type "second-generation target," if candidate information to be used in creating an answer to question Q can be directly searched for from the information type "first-generation target," that information is designated as candidate information to be used in creating an answer to question Q. On the other hand, if candidate information to be used in creating an answer to question Q cannot be directly searched for from the information type "first-generation target," a determination is made as to whether information to identify candidate information to be used in creating an answer to question Q can be searched for from the information type "second-generation target," and if it is determined that such information can be searched, information identified by linking to that information is designated as candidate information to be used in creating an answer to question Q. Note that, as in the former case, when a direct search is performed from the information type of first generation targets, it is, for example, when information highly similar to the first processing result can be searched from the information type of first generation targets. For example, it is when information highly similar to the first processing result can be searched from the information type of first generation targets more than information that can be searched from the information type of second generation targets. On the other hand, as in the latter case, when a search is performed from the information type of second generation targets, it is, for example, when information highly similar to the first processing result cannot be searched from the information type of first generation targets. For example, it is when information highly similar to the first processing result cannot be searched from the information type of first generation targets more than information that can be searched from the information type of second generation targets.

[0029] Furthermore, in FIG. 5, an example is given in which information is searched from the information type "first-generation target" and the information type "second-generation target" to determine candidate information to be used in creating an answer to question Q. However, only the information type "second-generation target" may be used. In other words, the search is not limited to searching both information that can be searched from the information type "first-generation target" and information that can be searched from the information type "second-generation target." In this case, candidate information to be used in creating an answer to question Q is not searched from the information type "first-generation target," but information for identifying candidate information to be used in creating an answer to question Q is searched from the information type "second-generation target." Then, information identified by linking to the information searched in this way is determined to be candidate information to be used in creating an answer to question Q.

[0030] 6A is an explanatory diagram illustrating the relationship between first-generation and second-generation objects according to an embodiment. In FIG. 6A, parent data D2 and parent data D3, etc., which are first-generation objects, are generated from source data D1; child data D4 and child data D5, which are second-generation objects, are generated from parent data D2; child data (grandchild data) D8 and child data (grandchild data) D9, which are second-generation objects, are generated from child data D4; and child data (grandchild data) D10 and child data (grandchild data) D11, which are second-generation objects, are generated from child data D5. Note that if source data D1 is the entire text of a company regulation, parent data D2 and parent data D3, etc., are part of the company regulation, child data D4 and child data 5 are further parts of that regulation, and child data (grandchild data) D8, child data (grandchild data) D9, child data (grandchild data) D10, and child data (grandchild data) D11 are further parts of that regulation.

[0031] FIG. 6B shows a specific example of FIG. 6A. In FIG. 6B, parent data D2 and parent data D3 in FIG. 6A correspond to parent chunks, child data D4 and child data 5 in FIG. 6A correspond to child chunk 1, and child data (grandchild data) D8, child data (grandchild data) D9, child data (grandchild data) D10, and child data (grandchild data) D11 in FIG. 6A correspond to child chunk 2. Note that in FIGS. 6A and 6B, child data (great-grandchild data from the perspective of the first generation) may be generated from each of child data (grandchild data) D8 to child data (grandchild data) 11, which are second-generation objects. The great-grandchild data generated in this way corresponds to child chunk 3 (not shown). In the following embodiment, with respect to the parent chunk, which is the first generation object, all child chunks after child chunk 1 (child chunk 2, child chunk 3, ...) are second-generation objects.

[0032] The information processing device 100 acquires the first processing result (step S201). That is, the information processing device 100 acquires information after converting the question Q. The information processing device 100 searches for information that is highly similar to the first processing result (step S202). At this time, the information processing device 100 performs a vector search together with the original question Q. For example, the information processing device 100 performs a search based on the similarity of vector features. Then, after searching for information from the information type of first-generation target and the information type of second-generation target, if the information processing device 100 has searched for information from the information type of second-generation target, the information processing device 100 identifies information linked to the searched information (information that falls under the information type of first-generation target) (step S203). In this case, for example, if a plurality of pieces of information that fall under the information type of second-generation target are searched, the information processing device 100 identifies pieces of information that fall under the information type of first-generation target by linking each piece of information. In step S203, the information processing device 100 searches for information from the information type of first generation and the information type of second generation, respectively, and if information is searched for from the information type of first generation, the information processing device 100 identifies the searched information (information that falls under the information type of first generation). In this case, for example, if a plurality of pieces of information that fall under the information type of first generation are searched for, the information processing device 100 identifies the plurality of pieces of information that fall under the information type of first generation.

[0033] The information processing device 100 may determine whether information highly similar to the first processing result can be found from the information type of first generation target as a result of searching for information from the information type of first generation target and the information type of second generation target, respectively. If the information processing device 100 determines that information highly similar to the first processing result can be found from the information type of first generation target, it may directly search for information that falls under the information type of first generation target from the information type of first generation target and identify the information thus found, and if it determines that information highly similar to the first processing result cannot be found from the information type of first generation target, it may search for information that falls under the information type of second generation target from the information type of second generation target and identify the information that falls under the information type of first generation target from the information thus found.

[0034] Furthermore, in steps S202 and S203, if there are multiple second-generation objects, the information processing device 100 may identify a generation object one level above the searched second-generation object. For example, using the examples of FIGS. 6A and 6B, if the searched second-generation object is child chunk 1, the information processing device 100 may identify a parent chunk linked to child chunk 1; if the searched second-generation object is child chunk 2, the information processing device 100 may identify child chunk 1 linked to child chunk 2; and if the searched second-generation object is child chunk 3, the information processing device 100 may identify child chunk 2 linked to child chunk 3. Then, the information processing device 100 may use the generation object identified in this way as a second processing result. In other words, it may be a candidate to be used in answering the question. 6A and 6B, if the searched second-generation object is child chunk 1, the information processing device 100 may identify a parent chunk linked to child chunk 1; if the searched second-generation object is child chunk 2, the information processing device 100 may identify a parent chunk linked to child chunk 1 which is linked to child chunk 2; if the searched second-generation object is child chunk 3, the information processing device 100 may identify a parent chunk linked to child chunk 1 which is linked to child chunk 2 which is linked to child chunk 3, and use these as the second processing result. Also, if a parent chunk is directly searched for, the information processing device 100 may identify the parent chunk searched for in this way and use this as the second processing result. In other words, the parent chunk identified in this way may be a candidate for use in answering the question.

[0035] In this case, the source data, the first-generation object, and the second-generation object may not be limited to text data and may be multimodal data such as image data. For example, the information processing device 100 may generate an image with reduced resolution as the first generation, an image with even lower resolution as the second generation, and generate each generation of image by pixel-dividing the image. Multimodal search can be achieved by vector-converting each generation of image along with text. For example, the information processing device 100 may recognize an image and an object captured in the image, convert the image into text using various models that output the recognized object in text format, and associate the text with text surrounding the image to generate metadata, thereby identifying the second processing result on a metadata-by-metadata basis. For example, when searching for text, the information processing device 100 may identify, as the second processing result, data of a generation object associated with the second-generation object, which is metadata associated with the searched text. Furthermore, for example, the information processing device 100 may directly search metadata and identify the searched metadata as the second processing result.

[0036] As a result, with conventional technology, even if there was a correct answer, if there was a lot of noise, search accuracy would decrease and appropriate searches would not be possible.However, with this technology, it is possible to search for highly relevant information using the embedding accuracy of the ``child,'' and by referencing the ``parent,'' it is possible to create comprehensive and detailed answers from a wide range of contexts.

[0037] FIG. 7 is a diagram illustrating an example of information processing (reranking) according to an embodiment. The information processing device 100 acquires a second processing result (step S301). That is, it acquires multiple search results searched based on a question Q. The information processing device 100 provides the question Q and the multiple search results to the generation AI along with a prompt for evaluating the relevance between the question Q and the multiple search results (step S302), and acquires an evaluation result from the generation AI (step S303). Note that this evaluation result is information for reranking the multiple search results that are the second processing result in order to create an answer to the question Q. That is, the information processing device 100 uses the evaluation result in step S303 to rerank the multiple search results that are the second processing result and extracts search results to be used in creating an answer to the question Q (step S304). Note that in this extraction process, the multiple search results are ranked in descending order of evaluation, thereby extracting a predetermined number of search results with the highest evaluations.

[0038] Then, the information processing device 100 provides the question Q and the extraction result together with a prompt for generating an answer to the question Q to the generation AI (step S305), and acquires the answer result from the generation AI (step S306).The information processing device 100 then provides the acquired answer result to the user U1 as the answer to the question Q (step S307).

[0039] As such, in Figure 7, the information processing device 100 evaluates the relevance of each of the multiple search results to question Q, extracts search results that are evaluated as having a high relevance, generates an answer to question Q using the extracted search results, and provides the generated answer to user U1.

[0040] FIG. 8 is an explanatory diagram for explaining reranking according to an embodiment. In FIG. 8, documents A to E are candidates for multiple search results based on question Q. Documents A to E are first-generation targets. Documents A to E are each evaluated for relevance to question Q by a generation AI, and top search results with high relevance are extracted. In FIG. 8, document D, which is evaluated for relevance the most, document B, which is evaluated for relevance the second most, and document C, which is evaluated for relevance the third most, are extracted. Then, an answer is created using the top search results with high relevance. In FIG. 8, an answer to question Q is created using document D, which is evaluated for relevance the most, document B, which is evaluated for relevance the second most, and document C, which is evaluated for relevance the third most.

[0041] In this case, documents A to E are not limited to text data, and may be multimodal data such as image data. Information processing device 100 may generate an answer to question Q based on image data or the like that is a plurality of search results. In this case, the final answer to question Q may be text data or image data.

[0042] As a result, information that has been evaluated as having a high degree of similarity in the vector search can be reevaluated based on the relevance to the question and narrowed down to the top-ranking information, making it possible to create highly accurate answers.

[0043] [4. Terminal Device Configuration] Next, the configuration of the terminal device 10 according to the embodiment will be described with reference to Fig. 9. Fig. 9 is a diagram showing an example of the configuration of the terminal device 10 according to the embodiment. As shown in Fig. 9, the terminal device 10 has a communication unit 11, an input unit 12, an output unit 13, and a control unit 14.

[0044] (Communications Department 11) The communication unit 11 is realized by, for example, a network interface card (NIC), etc. The communication unit 11 is connected to a predetermined network N by wire or wirelessly, and transmits and receives information to and from the information processing device 100, etc., via the predetermined network N.

[0045] (Input section 12) The input unit 12 accepts various operations from the user. In FIG. 4 and other figures, the input unit 12 accepts various operations from the user U1. For example, the input unit 12 may accept various operations from the user via a display screen using a touch panel function. The input unit 12 may also accept various operations from buttons provided on the terminal device 10 or a keyboard or mouse connected to the terminal device 10. For example, the input unit 12 accepts an operation for inputting a question.

[0046] (Output section 13) The output unit 13 is a display screen of a tablet terminal or the like realized by, for example, a liquid crystal display or an organic EL (Electro-Luminescence) display, and is a display device for displaying various information. For example, the output unit 13 displays information transmitted from the information processing device 100. For example, the output unit 13 displays an answer to a user's question generated by the information processing device 100.

[0047] (Control unit 14) The control unit 14 is, for example, a controller, and is realized by a CPU (Central Processing Unit) or an MPU (Micro Processing Unit) executing various programs stored in a storage device inside the terminal device 10 using a RAM (Random Access Memory) as a work area. For example, these various programs include application programs installed in the terminal device 10. For example, these various programs include application programs that display information transmitted from the information processing device 100 (such as answers to user questions). The control unit 14 is also realized by an integrated circuit, such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

[0048] As shown in FIG. 9, the control unit 14 has a receiving unit 141 and a transmitting unit 142, and realizes or executes the information processing operations described below.

[0049] (Receiving unit 141) The receiving unit 141 receives, for example, information transmitted from the information processing device 100. For example, the receiving unit 141 receives an answer to a user's question.

[0050] (Transmitter 142) The transmitting unit 142 transmits, for example, user operation information. For example, the transmitting unit 142 transmits a question received from the user. For example, the transmitting unit 142 transmits a question input, specified, or selected by the user.

[0051] 5. Configuration of Information Processing Device Next, the configuration of the information processing device 100 according to the embodiment will be described with reference to Fig. 10. Fig. 10 is a diagram showing an example of the configuration of the information processing device 100 according to the embodiment. As shown in Fig. 10, the information processing device 100 includes a communication unit 110, a storage unit 120, and a control unit 130. Note that the information processing device 100 may also include an input unit (e.g., a keyboard or a mouse) that accepts various operations from an administrator of the information processing device 100, and a display unit (e.g., a liquid crystal display) that displays various information.

[0052] (Communication unit 110) The communication unit 110 is realized by, for example, a NIC etc. The communication unit 110 is connected to a network N by wire or wirelessly, and transmits and receives information to and from the terminal device 10 etc. via the network N.

[0053] (Storage unit 120) The storage unit 120 is realized by, for example, a semiconductor memory element such as a RAM or a flash memory, or a storage device such as a hard disk or an optical disk. As shown in Fig. 10, the storage unit 120 has a first-generation target storage unit 121 and a second-generation target storage unit 122.

[0054] The first-generation object storage unit 121 stores first-generation objects. FIG. 11 shows an example of the first-generation object storage unit 121 according to the embodiment. The information stored in the first-generation object storage unit 121 is used, for example, to evaluate the degree of relevance to a user's question and to create an answer to the user's question. The information stored in the first-generation object storage unit 121 is also used, for example, to evaluate the degree of similarity to a user's question and to identify first-generation objects. As shown in FIG. 11, the first-generation object storage unit 121 has items such as "first-generation object ID," "source data ID," and "first-generation object."

[0055] "First generation target ID" indicates identification information for identifying the first generation target. "Source data ID" indicates identification information for identifying the source data linked to the first generation target of the target. "First generation target" indicates the first generation target. In the example shown in Figure 11, conceptual information such as "first generation target #1" and "first generation target #2" is stored in "first generation target," but in reality, text, images, etc. indicating the first generation target are stored.

[0056] The second-generation object storage unit 122 stores second-generation objects. Here, FIG. 12 shows an example of the second-generation object storage unit 122 according to the embodiment. The information stored in the second-generation object storage unit 122 is used, for example, to evaluate the similarity with a user's question, and also to identify first-generation objects. As shown in FIG. 12, the second-generation object storage unit 122 has items such as "second-generation object ID," "first-generation object ID (or second-generation object ID)," and "second-generation object."

[0057] "Second generation target ID" indicates identification information for identifying a second generation target. "First generation target ID (or second generation target ID)" indicates identification information for identifying a first generation target linked to the second generation target of the target. Also, if the second generation target of the target is a "grandchild," it indicates identification information for identifying the second generation target that is the "child." "Second generation target" indicates a second generation target. In the example shown in Figure 12, conceptual information such as "second generation target #1" and "second generation target #2" is stored in "second generation target," but in reality, text or images indicating the second generation target are stored.

[0058] (control unit 130) The control unit 130 is a controller, and is realized by, for example, a CPU or an MPU executing various programs stored in a storage device inside the information processing device 100 using RAM as a work area. The control unit 130 is also realized by, for example, an integrated circuit such as an ASIC or an FPGA.

[0059] 10, the control unit 130 has an acquisition unit 131, a conversion unit 132, a search unit 133, an identification unit 134, an evaluation unit 135, an extraction unit 136, a generation unit 137, and a provision unit 138, and realizes or executes the information processing action described below. Note that the internal configuration of the control unit 130 is not limited to the configuration shown in FIG. 10, and may be any other configuration as long as it performs the information processing described below.

[0060] (Acquisition part 131) The acquiring unit 131 acquires various pieces of information from the storage unit 120. The acquiring unit 131 also stores the acquired various pieces of information in the storage unit 120.

[0061] The acquisition unit 131 acquires various pieces of information from an external information processing device. The acquisition unit 131 acquires various pieces of information from other information processing devices such as the terminal device 10.

[0062] The acquiring unit 131 acquires, for example, a question from a user. For example, the acquiring unit 131 acquires a question that is input, specified, or selected by a user.

[0063] (Conversion unit 132) The conversion unit 132, for example, converts a user's question into specific information. For example, the conversion unit 132 uses a generation AI to convert a user's question into specific information in order to create an answer to the user's question. That is, for example, the conversion unit 132 converts the user's question into information for creating an answer to the user's question.

[0064] The conversion unit 132 converts the information into information for which the similarity is determined in order to narrow down the target (corresponding to the search target) that will be the basis for creating an answer. For example, the conversion unit 132 converts the information into information for which the similarity is determined by vector search together with the user's question.

[0065] The conversion unit 132 performs the conversion process when the user's question satisfies a predetermined condition. For example, the conversion unit 132 performs the conversion process when the user's question is not too short (when the question is equal to or longer than a predetermined number of characters).

[0066] The conversion unit 132 converts, for example, an example of information that embodies a user's question into information that is a tentative answer to the question. Furthermore, the conversion unit 132 converts, for example, an example of information that embodies a user's question into information that is one or more questions similar to the question.

[0067] The conversion unit 132 converts the user's question into specific information, for example, by providing the generation AI with an instruction sentence indicating an instruction to convert the question. Also, the conversion unit 132 converts the user's question into specific information, for example, by providing the generation AI with an instruction sentence expressing at least one of the role, task content, task purpose, premise, and constraint condition in natural language.

[0068] (Search section 133) The search unit 133 performs a search based on, for example, information converted by the conversion unit 132. For example, the search unit 133 searches for information that has a high degree of similarity (cosine similarity) based on a vector search with the information converted by the conversion unit 132. For example, the search unit 133 searches for information that falls under the information type of second-generation target, which has been generated based on information that falls under the information type of first-generation target. For example, the search unit 133 searches for information that falls under the information type of second-generation target, which has been generated by dividing information that falls under the information type of first-generation target into predetermined sentence units, or information that falls under the information type of second-generation target, which has been generated by dividing and dividing information that falls under the information type of first-generation target into sentence units so that the number of characters is equal to or less than the maximum number. Also, for example, the search unit 133 searches for information that falls under the information type of first-generation target. For example, the search unit 133 searches for information from the information type "first generation target" and the information type "second generation target," respectively, and determines whether information highly similar to the first processing result can be found from the information type "first generation target." If it is determined that information highly similar to the first processing result can be found from the information type "first generation target," the search unit 133 searches for information corresponding to the first generation target. Note that, at this time, for example, if the search unit 133 determines that information highly similar to the first processing result cannot be found from the information type "first generation target," the search unit 133 searches for information corresponding to the second generation target. Furthermore, for example, the search unit 133 searches for multiple pieces of information (information corresponding to a predetermined generation target) that are highly similar based on a vector search with the information converted by the conversion unit 132. For example, the search unit 133 searches for a predetermined number of pieces of information (information corresponding to a predetermined generation target) that have the highest similarity.

[0069] (Specific Section 134) The identification unit 134 identifies information corresponding to the information type of first-generation targets that will be used as a basis for creating an answer to a question, based on, for example, information corresponding to the information type of second-generation targets searched by the search unit 133. For example, the identification unit 134 identifies information corresponding to the information type of first-generation targets that is linked to information corresponding to the information type of second-generation targets searched by the search unit 133. For example, the identification unit 134 identifies each piece of information corresponding to the information type of first-generation targets that is linked to multiple pieces of information corresponding to the information type of second-generation targets searched by the search unit 133. Furthermore, the identification unit 134 identifies the information corresponding to the information type of first-generation targets searched by the search unit 133 as information that will be used as a basis for creating an answer to a question. For example, the identification unit 134 identifies each piece of information corresponding to the information type of first-generation targets searched by the search unit 133.

[0070] (Evaluation Section 135) The evaluation unit 135 evaluates, for example, the relevance between the first-generation objects (corresponding to the search results) identified by the identification unit 134 and the user's question. For example, the evaluation unit 135 evaluates the relevance between each of the multiple first-generation objects and the user's question. For example, the evaluation unit 135 performs the evaluation by providing the generation AI with an instruction sentence indicating an instruction to evaluate the relevance between each of the multiple first-generation objects and the user's question. For example, the evaluation unit 135 performs the evaluation by providing the generation AI with the multiple first-generation objects and the user's question together with an instruction sentence indicating an instruction to evaluate the relevance between each of the multiple first-generation objects and the user's question.

[0071] (Extraction part 136) The extraction unit 136 extracts, for example, first-generation objects whose relevance has been highly evaluated by the evaluation unit 135 from among the first-generation objects identified by the identification unit 134. For example, the extraction unit 136 extracts a predetermined number of first-generation objects with high relevance by ranking the first-generation objects identified by the identification unit 134 in descending order of relevance.

[0072] (Generation unit 137) The generation unit 137 generates an answer to the user's question, for example, by using one or more first-generation objects extracted by the extraction unit 136. For example, the generation unit 137 generates an answer to the user's question by providing the generation AI with an instruction sentence indicating an instruction to generate an answer to the question using the first-generation objects. For example, the generation unit 137 generates an answer to the user's question by providing the generation AI with one or more first-generation objects and the user's question, along with an instruction sentence indicating an instruction to generate an answer to the question using the first-generation objects.

[0073] (Provider 138) The providing unit 138 provides the answer generated by the generating unit 137 to the user as the answer to the user's question, for example.

[0074] [6. Information Processing Flow] Next, the procedure of information processing by the information processing system 1 according to the embodiment will be described with reference to Fig. 13 to Fig. 15. Fig. 13 is a flowchart showing the procedure of question conversion processing according to the embodiment, Fig. 14 is a flowchart showing the procedure of search refinement processing according to the embodiment, and Fig. 15 is a flowchart showing the procedure of reranking processing according to the embodiment.

[0075] As shown in FIG. 13, the information processing device 100 acquires a user's question (step S401). The information processing device 100 determines whether the user's question satisfies a predetermined condition (step S402). If the information processing device 100 determines that the user's question satisfies the predetermined condition (step S402; YES), the information processing device 100 provides the question to the generation AI together with an instruction sentence indicating an instruction to convert the question (step S403). Then, the information processing device 100 converts the user's question into specific information (step S404). On the other hand, if the information processing device 100 determines that the user's question does not satisfy the predetermined condition (step S402; NO), the information processing device 100 terminates information processing. In this case, the information processing device 100 may notify the user to request the user to ask the question again (the content of the question may be the same) or to change the question.

[0076] 14, the information processing device 100 acquires information that embodies a user's question (step S501). The information processing device 100 compares the feature amounts of the information that embodies the user's question with the feature amounts of candidate first-generation objects and candidate second-generation objects to search for multiple first-generation objects or second-generation objects that have a high similarity to the information that embodies the user's question (step S502). The information processing device 100 identifies the first-generation objects that have been found or first-generation objects that are linked to the second-generation objects that have been found, thereby identifying information that will serve as a basis for creating an answer to the user's question (step S503).

[0077] In steps S502 and S503, the information processing device 100 may determine the similarity between the information specifying the user's question and the candidate first-generation objects by comparing the features of the information specifying the user's question and the candidate first-generation objects, and if a candidate first-generation object with a high similarity exists, may search for multiple first-generation objects with a high similarity and identify the searched multiple first-generation objects to identify information that will serve as a basis for creating an answer to the user's question. In addition, in steps S502 and S503, the information processing device 100 may determine the similarity between the information specifying the user's question and the candidate first-generation objects by comparing the features of the information specifying the user's question and the candidate first-generation objects, and if a candidate first-generation object with a high similarity does not exist, may search for multiple second-generation objects with a high similarity and identify first-generation objects linked to the searched multiple second-generation objects to identify information that will serve as a basis for creating an answer to the user's question. At this time, if there are no candidates for second-generation objects with high similarity, the information processing device 100 may notify the user that it cannot provide a highly accurate answer result or may notify the user that it requests that the question be changed (obviously, the notification is not limited to these examples). Also, in steps S502 and S503, the information processing device 100 may determine both the similarity with the first-generation object and the similarity with the second-generation object, search for multiple generation objects with higher similarity, and identify multiple first-generation objects based on the search results, thereby identifying information that will be the basis for creating an answer to the user's question.

[0078] As shown in FIG. 15, the information processing device 100 acquires a plurality of first-generation objects that satisfy a predetermined condition (step S601). The information processing device 100 provides the plurality of first-generation objects and the question to the generation AI along with an instruction sentence indicating an instruction to evaluate the relevance between each of the plurality of first-generation objects and the user's question (step S602). The information processing device 100 evaluates the relevance between each of the plurality of first-generation objects and the user's question, and reranks the first-generation objects in descending order of relevance according to the relevance (step S603). The information processing device 100 extracts a predetermined number of first-generation objects with the highest relevance based on the reranking results (step S604). Then, the information processing device 100 generates an answer to the user's question based on the extraction result (step S605).

[0079] [7. Effects] As described above, the information processing device 100 according to the embodiment includes the search unit 133 and the identification unit 134. The search unit 133 searches for second-generation objects that are generation objects that have a high degree of similarity to information based on a user's question and that are generated based on first-generation objects. The identification unit 134 identifies first-generation objects that will serve as a basis for creating an answer to the question, based on the second-generation objects searched for by the search unit 133.

[0080] As a result, the information processing device 100 according to the embodiment can, for example, promote improvement in search performance for generating appropriate answers to user questions. Furthermore, the information processing device 100 according to the embodiment can, for example, search for highly relevant information, and can generate comprehensive and detailed answers from a wide range of contexts by referring to the "parent."

[0081] In addition, the search unit 133 searches for second-generation targets that are highly similar to the information generated by providing the question to a generation AI that is capable of generating a provisional answer to a specified question or generating one or more questions similar to the specified question.

[0082] As a result, the information processing apparatus 100 according to the embodiment can perform a more accurate search using, for example, the converted question, thereby promoting improvement in search performance.

[0083] In addition, the search unit 133 searches for second-generation objects based on information generated by providing the generation AI with prompts and questions for generating information for determining similarity along with questions to search for second-generation objects.

[0084] This allows the information processing apparatus 100 according to the embodiment to search for, for example, highly relevant information, thereby enabling more accurate answer generation.

[0085] The second generation object is a generation object generated by dividing the first generation object into predetermined sentence units.

[0086] As a result, the information processing apparatus 100 according to the embodiment can promote improvement of search performance by, for example, appropriate chunk division in sentence units.

[0087] The second generation object is a generation object generated by dividing the first generation object into sentence units so that the number of characters is equal to or less than the maximum number.

[0088] As a result, the information processing apparatus 100 according to the embodiment can promote improvement of search performance by dividing data into chunks appropriately based on the number of characters, for example.

[0089] The first generation object is a generation object generated from predetermined source data on the premise that it will be used to answer questions.

[0090] As a result, the information processing apparatus 100 according to the embodiment can appropriately use chunks generated on the assumption that they will be used to answer a question, for example, to generate an answer.

[0091] The predetermined source data is company regulations created for company use.

[0092] As a result, the information processing apparatus 100 according to the embodiment can promote improvement in search performance for creating appropriate answers to user questions about company regulations, for example.

[0093] [8. Hardware Configuration] The information processing device 100 according to the embodiment described above is realized, for example, by a computer 1000 configured as shown in Fig. 16. Fig. 16 is a hardware configuration diagram showing an example of a computer that realizes the functions of the information processing device 100. The computer 1000 has a CPU 1100, a RAM 1200, a ROM 1300, an HDD 1400, a communication interface (I / F) 1500, an input / output interface (I / F) 1600, and a media interface (I / F) 1700.

[0094] The CPU 1100 operates and controls each unit based on programs stored in the ROM 1300 or the HDD 1400. The ROM 1300 stores a boot program executed by the CPU 1100 when the computer 1000 starts up, programs that depend on the hardware of the computer 1000, and the like.

[0095] The HDD 1400 stores programs executed by the CPU 1100, data used by these programs, etc. The communication interface 1500 acquires data from other devices via a predetermined communication network and sends it to the CPU 1100, and transmits data generated by the CPU 1100 to other devices via the predetermined communication network.

[0096] The CPU 1100 controls output devices such as a display and a printer, and input devices such as a keyboard and a mouse, via the input / output interface 1600. The CPU 1100 acquires data from the input devices via the input / output interface 1600. The CPU 1100 also outputs generated data to the output devices via the input / output interface 1600.

[0097] Media interface 1700 reads a program or data stored in recording medium 1800 and provides it to CPU 1100 via RAM 1200. CPU 1100 loads the program or data from recording medium 1800 onto RAM 1200 via media interface 1700 and executes the loaded program. Recording medium 1800 is, for example, an optical recording medium such as a DVD (Digital Versatile Disc) or a PD (Phase Change Rewritable Disc), a magneto-optical recording medium such as an MO (Magneto-Optical disk), a tape medium, a magnetic recording medium, or a semiconductor memory.

[0098] For example, when the computer 1000 functions as the information processing device 100 according to the embodiment, the CPU 1100 of the computer 1000 executes programs loaded onto the RAM 1200 to realize the functions of the control unit 130. The CPU 1100 of the computer 1000 reads and executes these programs from the recording medium 1800, but as another example, the CPU 1100 may obtain these programs from another device via a predetermined communication network.

[0099] [9. Other] Furthermore, among the processes described in the above embodiments, all or part of the processes described as being performed automatically can be performed manually, or all or part of the processes described as being performed manually can be performed automatically using a known method. In addition, the information including the processing procedures, specific names, various data, and parameters shown in the above documents and drawings can be changed as desired unless otherwise specified. For example, the various information shown in each drawing is not limited to the information shown in the drawings.

[0100] Furthermore, the components of each device shown in the figure are conceptual functional components and do not necessarily have to be physically configured as shown in the figure. In other words, the specific form of distribution and integration of each device is not limited to that shown in the figure, and all or part of them can be functionally or physically distributed and integrated in any unit depending on various loads, usage conditions, etc.

[0101] Furthermore, the above-described embodiments can be combined as appropriate within the scope of not causing any contradiction in the processing content.

[0102] Although some of the embodiments of the present application have been described in detail above with reference to the drawings, these are merely examples, and the present invention can be implemented in other forms that include the embodiments described in the Disclosure of the Invention section and that have undergone various modifications and improvements based on the knowledge of those skilled in the art.

[0103] Furthermore, the above-mentioned "section, module, unit" can be read as "means" or "circuit," etc. For example, an acquisition unit can be read as an acquisition means or an acquisition circuit. [Explanation of symbols]

[0104] 1. Information Processing Systems 10 Terminal Equipment 11 Communications Department 12 Input section 13 Output section 14 Control Unit 100 Information processing device 110 Communications Department 120 Storage section 121 First Generation Target Memory Unit 122 Second Generation Target Memory Unit 130 Control Unit 131 Acquisition Department 132 Conversion Unit 133 Search Department 134 Specific part 135 Evaluation Department 136 Extraction part 137 Generation part 138 Provision Department 141 Receiving unit 142 Transmitter N Network

Claims

1. a search unit that searches for second-generation objects that are generation objects that have a high similarity to information based on a user's question and are generated based on the first-generation objects; an identification unit that identifies the first-generation object that will be used as a basis for creating an answer to the question based on the second-generation object searched for by the search unit; An information processing device comprising:

2. The search unit The question is provided to a generation AI capable of generating a provisional answer to a predetermined question or generating one or more questions similar to the predetermined question, thereby searching for the second-generation object having a high similarity to the generated information.

2. The information processing apparatus according to claim 1, wherein:

3. The search unit and searching for the second-generation object based on the information generated by providing the prompt for generating the information for which the similarity is determined together with the question to the generation AI in order to search for the second-generation object, and the question.

3. The information processing apparatus according to claim 2, wherein:

4. The second generation object is a generation object generated by dividing the first generation object into predetermined sentence units.

2. The information processing apparatus according to claim 1, wherein:

5. The second generation object is a generation object generated by dividing the first generation object into sentence units so that the number of characters is equal to or less than the maximum number.

2. The information processing apparatus according to claim 1, wherein:

6. The first generation object is a generation object generated from predetermined source data on the premise that it will be used to answer questions.

2. The information processing apparatus according to claim 1, wherein:

7. The predetermined source data is a company rule created for the company.

7. The information processing apparatus according to claim 6,

8. 1. A computer-implemented information processing method, comprising: a search step of searching for a second generation object that is a generation object having a high similarity to information based on a user's question and that is generated based on the first generation object; an identifying step of identifying the first-generation object that will be used as a basis for creating an answer to the question, based on the second-generation object searched for by the searching step; An information processing method comprising:

9. a search procedure for searching for a second generation object that is a generation object having a high similarity to information based on a user's question and that is generated based on the first generation object; an identification step of identifying the first-generation object that will be used as a basis for creating an answer to the question, based on the second-generation object searched by the search step; An information processing program characterized by causing a computer to execute the above.

Images