Information processing system and information processing method

Abstract

Provided is a novel information processing system that is excellent in convenience, usefulness and reliability. This information processing system includes: a first component for receiving, from a user, a question and a reply described in a natural language, and providing a response and a report to the user; a second component for generating an inquiry sentence and an answer using a large-scale language model; and a third component for creating a response that includes the inquiry sentence and a search results list, and a report that includes the answer and a search results list. The third component uses the question to find relevant information from a search index, calculates a search results list and degrees of similarity, and refers to background information adopted on the basis of the degrees of similarity to generate an instruction sentence for obtaining an answer to the question.

Description

Information processing system and information processing method

[0001] One embodiment of the present invention relates to an information processing system, an information processing method, or a semiconductor device.

[0002] Note that one embodiment of the present invention is not limited to the above technical field. The technical field of one embodiment of the invention disclosed in this specification relates to an object, a method, or a manufacturing method. Alternatively, one embodiment of the present invention relates to a process, a machine, a manufacture, or a composition of matter. Therefore, more specifically, examples of the technical field of one embodiment of the present invention disclosed in this specification include a data processing device, a semiconductor device, a memory device, a driving method thereof, or a manufacturing method thereof.

[0003] In recent years, the development of language models using neural networks has been actively carried out, and large-scale language models (LLMs) in particular have attracted attention. A large-scale language model is a natural language processing model trained using a large amount of data. A large-scale language model can realize a dialogue model that responds to user instructions, for example. Non-Patent Document 1 discloses GPT-4 (Generative Pre-trained Transformer 4) (registered trademark) as a large-scale language model, and also discloses ChatGPT as a dialogue model.

[0004] The use of large-scale language models has significantly increased the capabilities of natural language processing models. However, as language models become larger, it is difficult to incorporate and operate language models in-house due to the equipment and cost involved. Therefore, one way to use language models is to use external services that provide language models.

[0005] Summary of ChatGPT / GPT-4 Research and Perspective Towards the Future of Large Language Models, Yiheng Liu et al. (Submitted on 4 Apr 2023, [online], Internet <URL: https: / / arxiv.org / abs / 2304.01852>

[0006] An object of one embodiment of the present invention is to provide a novel information processing system with excellent convenience, usefulness, or reliability, or to provide a novel information processing method with excellent convenience, usefulness, or reliability, or to provide a novel information processing system, a novel information processing method, or a novel semiconductor device.

[0007] Note that the description of these problems does not preclude the existence of other problems. Note that one embodiment of the present invention does not necessarily solve all of these problems. Note that problems other than these will become apparent from the description of the specification, drawings, claims, etc., and it is possible to extract other problems from the description of the specification, drawings, claims, etc.

[0008] (1) One aspect of the present invention is an information processing system having a first component, a second component, and a third component.

[0009] The first component has a function of accepting and transmitting questions and replies, which are written in natural language, to the third component, and a function of accepting and providing responses and reports.

[0010] The second component has a function of accepting the instruction sentence and sending a query sentence and an answer to the third component. The second component also has a function of performing processing using a large-scale language model. The large-scale language model has a function of determining whether the question is clear and appropriate according to the instruction sentence and a function of generating the query sentence and an answer.

[0011] The third component comprises a first subcomponent, a second subcomponent and a third subcomponent.

[0012] The first subcomponent includes a function for supplementing and updating a question using responses and replies, a search engine, and a search index. The search engine uses the question to find related information from the search index, calculates similarity scores, and creates a search result list. The search result list includes related information headings sorted in descending order of similarity, where the similarity score represents the strength of the relationship between the question and the related information. The first subcomponent also includes a function for incorporating background information from the related information based on the similarity scores.

[0013] The second subcomponent has a function of creating and sending to the second component an instruction statement, the instruction statement including a question, background information, first instructions for generating an answer to the question by referring to the background information, and second instructions for generating a query statement to generate a more detailed answer, withholding generation of the answer as necessary. The query statement also includes a prompt for follow-up and a prompt for confirmation regarding the question.

[0014] The third subcomponent is responsible for generating and sending responses and reports to the first component, where the responses include queries and search result lists, and the reports include answers and search result lists.

[0015] This allows documents with different attributes to be registered in a search index, enabling documents with different attributes to be treated as related information across a wide range of documents. Furthermore, since a huge number of documents can be registered in a search index, background information can be adopted across a wide range of documents. Furthermore, background information can be adopted across a wide range of documents, for example, from official documents, patent publications, patent gazettes, books, literature, manuals, and private documents. Furthermore, background information can be adopted from confidential documents, confidential documents, and other documents. Furthermore, background information can be adopted from related information based on a similarity that indicates the strength of the relationship between a question and related information. Generation of an answer can be postponed when a question is unclear. Generation of an answer can be postponed when a question is inappropriate. Furthermore, unnecessary processing using a large-scale language model can be avoided. Furthermore, questions can be improved to be clear and appropriate through dialogue using responses and replies. Furthermore, dialogue using responses and replies can, for example, complement a user's linguistic expression ability. Furthermore, related information can be obtained using clear and appropriate questions. Furthermore, noise contained in related information can be reduced. As a result, a novel information processing system with excellent convenience, usefulness, and reliability can be provided.

[0016] (2) Another aspect of the present invention is the information processing system, wherein the first component has a function of receiving a stop command and transmitting the stop command to the third component. Note that the third component has a function of transmitting the stop command to the second component, and the second component has a function of receiving the stop command and terminating processing using the large-scale language model.

[0017] This allows the processing using a large-scale language model to be stopped, thereby eliminating wasteful processing. For example, it is possible to eliminate wasteful processing time and power consumption. As a result, it is possible to provide a novel information processing system that is highly convenient, useful, and reliable.

[0018] (3) In accordance with another aspect of the present invention, the first component has a function of receiving the selection information and transmitting the selection information to the third component, and the background information includes related information adopted based on the selection information instead of the similarity.

[0019] As a result, even if unintended background information is adopted, it can be re-adopted using the selected information. Furthermore, even if unintended background information is adopted due to, for example, a user's linguistic expression ability, it can be re-adopted using the selected information. Furthermore, even if unintended background information is adopted due to, for example, a bias in information registered in a search index, it can be re-adopted using the selected information. As a result, a novel information processing system that is highly convenient, useful, and reliable can be provided.

[0020] (4) One aspect of the present invention is an information processing method including first to fourteenth steps.

[0021] In the first step, a first component accepts a query and sends it to a second component.

[0022] In a second step, a second component receives the question and generates a list of search results. The second component includes a search engine and a search index. The search engine uses the question to find related information from the search index and calculates a similarity score. The list of search results includes headlines of related information sorted in descending order of similarity. The similarity score represents the strength of the relationship between the question and the related information.

[0023] In the third step, the second component employs background information from the related information based on similarity.

[0024] In a fourth step, the second component creates and sends a response to the first component, the response including the list of search results.

[0025] In the fifth step, the first component provides a response.

[0026] In a sixth step, the second component creates and sends to the third component an instruction statement, the instruction statement including the question, the background information, and instructions for generating an answer to the question by referencing the background information.

[0027] In a seventh step, the third component receives the instruction sentence and starts generating an answer using the large-scale language model, where the third component has a function of performing processing using the large-scale language model, and the large-scale language model has a function of generating an answer according to the instruction sentence.

[0028] In the eighth step, if the first component receives the selection before the third component has completed generating the answer, the first component passes the selection to the second component and the process proceeds to the ninth step, otherwise the process proceeds to the twelfth step.

[0029] In the ninth step, the second component passes the abort command to the third component and re-adopts background information from the relevant information based on the selection information.

[0030] In the tenth step, the third component accepts the stop command and stops generating the answer.

[0031] In an eleventh step, the second component updates the response and passes it to the first component, and the process proceeds to the fifth step, where the updated response includes the search result list and selection information.

[0032] In a twelfth step, the third component sends a response to the second component.

[0033] In a thirteenth step, the second component accepts the answer and generates and sends a report to the first component, the report including the answer and the search results list.

[0034] In a fourteenth step, the first component provides a report.

[0035] This allows documents with different attributes to be registered in a search index, making it possible to treat documents with different attributes as related information across a wide range of documents. Furthermore, since a huge number of documents can be registered in a search index, background information can be adopted across a wide range of documents. Furthermore, background information can be adopted across a wide range of documents, for example, from official documents, patent publications, patent gazettes, books, literature, manuals, private documents, etc. Furthermore, background information can be adopted from confidential documents, confidential documents, etc. Furthermore, background information can be adopted from related information based on a similarity that indicates the strength of the relationship between a question and related information. Even if unintended background information is adopted, it can be re-adopted using selected information. Even if unintended background information is adopted due to, for example, a user's linguistic expression ability, it can be re-adopted using selected information. Even if unintended background information is adopted due to, for example, a bias in the information registered in the search index, it can be re-adopted using selected information. Furthermore, while determining whether the adopted background information is appropriate, the large-scale language model can generate an answer. Furthermore, time is not wasted determining whether the adopted background information is appropriate. As a result, a novel information processing method that is highly convenient, useful, and reliable can be provided.

[0036] (5) One aspect of the present invention is an information processing method including first to twenty-first steps.

[0037] In the first step, a first component accepts a query and sends it to a second component.

[0038] In a second step, a second component receives the question and generates a list of search results. The second component includes a search engine and a search index. The search engine uses the question to find related information from the search index and calculates a similarity score. The list of search results includes headlines of related information sorted in descending order of similarity. The similarity score represents the strength of the relationship between the question and the related information.

[0039] In the third step, the second component employs background information from the related information based on similarity.

[0040] In a fourth step, the second component creates and sends a response to the first component, the response including the list of search results.

[0041] In the fifth step, the first component provides a response.

[0042] In a sixth step, the second component creates and sends to the third component an instruction statement, the instruction statement including a question, background information, first instructions for generating an answer to the question by referring to the background information, and second instructions for generating a query statement to generate a more detailed answer, withholding generation of the answer as necessary, and the query statement including a prompt for follow-up and a prompt for confirmation regarding the question.

[0043] In the seventh step, the third component receives the instruction sentence and considers the question using the large-scale language model. The third component has a function of performing processing using the large-scale language model. The large-scale language model has a function of determining whether the question is clear and appropriate according to the instruction sentence, and a function of generating a query sentence and an answer.

[0044] In the eighth step, if the large-scale language model determines that the question is not clear or appropriate, the process proceeds to the ninth step, and if the large-scale language model determines that the question is clear and appropriate, the process proceeds to the fourteenth step.

[0045] In the ninth step, the third component generates a query statement and sends it to the second component.

[0046] In a tenth step, the second component updates the response and sends it to the first component, where the response in the tenth step includes the query statement and the list of search results.

[0047] In an eleventh step, the first component provides a response.

[0048] In a twelfth step, the first component accepts the reply and sends it to the second component.

[0049] In the thirteenth step, the second component accepts the response, updates the question, and advances the process to the second step, where the updated question includes the response and the reply.

[0050] In the fourteenth step, the third component begins generating answers using the large-scale language model.

[0051] In step 15, if the first component receives the selection information before the third component completes generating the answer, the first component passes the selection information to the second component and the process proceeds to step 16; otherwise, the process proceeds to step 19.

[0052] In a sixteenth step, the second component passes the abort command to the third component and re-adopts background information from the relevant information based on the selection information.

[0053] In a seventeenth step, the third component accepts the stop command and stops generating the answer.

[0054] In the eighteenth step, the second component updates the response and passes it to the first component, and the process proceeds to the fifth step. Note that the response in the eighteenth step includes the list of search results and selection information.

[0055] In a nineteenth step, the third component sends a response to the second component.

[0056] In a twentieth step, the second component accepts the answer and generates and sends a report to the first component, the report including the answer and the list of search results.

[0057] In a twenty-first step, the first component provides a report.

[0058] This makes it possible to withhold the generation of an answer when the question is unclear. Also, it makes it possible to withhold the generation of an answer when the question is inappropriate. Also, it makes it possible to eliminate unnecessary processing using a large-scale language model. Also, it makes it possible to improve the question to be clear and appropriate through a dialogue using responses and replies. Also, it makes it possible to supplement the user's linguistic expression ability through a dialogue using responses and replies. Also, it makes it possible to obtain related information using clear and appropriate questions. Also, it makes it possible to reduce noise contained in the related information. As a result, it is possible to provide a novel information processing method that is highly convenient, useful, and reliable.

[0059] One embodiment of the present invention can provide a novel information processing system with excellent convenience, usefulness, or reliability, or a novel information processing method with excellent convenience, usefulness, or reliability, or a novel information processing system, a novel information processing method, or a novel semiconductor device.

[0060] Note that the description of these effects does not preclude the existence of other effects. Note that one embodiment of the present invention does not necessarily have all of these effects. Note that effects other than these will become apparent from the description in the specification, drawings, claims, etc., and it is possible to extract other effects from the description in the specification, drawings, claims, etc.

[0061] FIG. 1 is a diagram illustrating the configuration of an information processing system according to an embodiment. FIG. 2 is a diagram illustrating the configuration of components used in the information processing system according to an embodiment. FIG. 3 is a diagram illustrating the configuration of an instruction statement used in the information processing system according to an embodiment. FIGS. 4A and 4B are diagrams illustrating the configuration of a response and a report used in the information processing system according to an embodiment. FIG. 5 is a diagram illustrating the configuration of an information processing device used in the information processing system according to an embodiment. FIG. 6 is a diagram illustrating an information processing method according to an embodiment. FIG. 7 is a diagram illustrating the information processing method according to an embodiment.

[0062] An information processing system according to one aspect of the present invention includes a first component, a second component, and a third component.

[0063] The first component has the functionality to accept and send questions and replies to the third component, the questions and replies being written in natural language, and the first component has the functionality to accept and provide responses and reports.

[0064] The second component has a function of accepting the instruction sentence and sending a query sentence and an answer to the third component, and the second component has a function of processing using a large-scale language model, which has a function of determining whether the question is clear and appropriate according to the instruction sentence and a function of generating the query sentence and an answer.

[0065] The third component comprises a first subcomponent, a second subcomponent and a third subcomponent.

[0066] The first subcomponent includes a function for supplementing and updating the question with responses and replies, a search engine, and a search index. The search engine uses the question to find related information from the search index, calculates similarity scores, and creates a search result list. The search result list includes headings of related information sorted in descending order of similarity, where the similarity score represents the strength of the association between the question and the related information. The first subcomponent also includes a function for incorporating background information from the related information based on the similarity scores.

[0067] The second subcomponent has a function of creating and sending to the second component an instruction statement, the instruction statement including a question, background information, first instructions for generating an answer to the question by referring to the background information, and second instructions for generating a query statement to generate a more detailed answer, withholding generation of the answer as necessary, the query statement including a prompt for follow-up regarding the question and a prompt for confirmation regarding the question.

[0068] The third subcomponent has the functionality to generate and send responses and reports to the first component, where the responses include a query statement and a search result list, and the reports include an answer and a search result list.

[0069] This allows documents with different attributes to be registered in a search index, enabling documents with different attributes to be treated as related information across a wide range of documents. Furthermore, since a huge number of documents can be registered in a search index, background information can be adopted across a wide range of documents. Furthermore, background information can be adopted across a wide range of documents, for example, from official documents, patent publications, patent gazettes, books, literature, manuals, and private documents. Furthermore, background information can be adopted from confidential documents, confidential documents, and other documents. Furthermore, background information can be adopted from related information based on a similarity that indicates the strength of the relationship between a question and related information. Generation of an answer can be postponed when a question is unclear. Generation of an answer can be postponed when a question is inappropriate. Furthermore, unnecessary processing using a large-scale language model can be avoided. Furthermore, questions can be improved to be clear and appropriate through dialogue using responses and replies. Furthermore, dialogue using responses and replies can, for example, complement a user's linguistic expression ability. Furthermore, related information can be obtained using clear and appropriate questions. Furthermore, noise contained in related information can be reduced. As a result, a novel information processing system with excellent convenience, usefulness, and reliability can be provided.

[0070] The embodiments will be described in detail with reference to the drawings. However, the present invention is not limited to the following description, and it will be readily understood by those skilled in the art that various changes in form and details can be made without departing from the spirit and scope of the present invention. Therefore, the present invention should not be interpreted as being limited to the description of the embodiments shown below. In the configuration of the invention described below, the same parts or parts having similar functions will be denoted by the same reference numerals in different drawings, and repeated explanations will be omitted.

[0071] In the drawings accompanying this specification, components are classified by function and shown as block diagrams that are independent of each other, but in reality, it is difficult to completely separate components by function, and one component may be involved in multiple functions.

[0072] Embodiment 1 In this embodiment, an information processing system according to one embodiment of the present invention will be described with reference to FIGS.

[0073] FIG. 1 is a diagram illustrating a configuration of an information processing system according to one embodiment of the present invention.

[0074] FIG. 2 is a diagram illustrating the configuration of components of an information processing system according to an embodiment of the present invention.

[0075] FIG. 3 is a schematic diagram illustrating the configuration of a directive used by an information processing system according to an embodiment of the present invention.

[0076] FIG. 4A is a schematic diagram illustrating the configuration of a response provided by an information processing system according to one aspect of the present invention, and FIG. 4B is a schematic diagram illustrating the configuration of a report.

[0077] FIG. 5 is a block diagram illustrating a configuration of an information processing device that can be used in an information processing system of one embodiment of the present invention.

[0078] <Configuration Example 1 of Information Processing System> An information processing system according to one aspect of the present invention described in this embodiment includes a component 30, a component 20, and a component 21 (see FIG. 1).

[0079] <<Configuration Example of Component 30>> The component 30 has a function of receiving a question Qre and a reply Rep from a user and transmitting them to the component 21. The question Qre and the reply Rep are written in natural language.

[0080] The component 30 also has a function of receiving responses Res and reports Rpt from the component 21 and providing them to the user.

[0081] For example, information specifying the range of topics to which the question Qre belongs can be included in the question Qre. Specifically, "politics," "economy," "culture," "science," "society," "incidents," etc. can be used as information specifying the range of topics. Furthermore, "natural science," "technology / engineering," "biology / agriculture," "medicine / pharmacy / psychology," etc. can be used as information specifying the range of topics. Furthermore, "materials engineering," "architecture / architectural engineering," "transportation engineering," "energy engineering," "information / communications / electronic engineering," "food engineering," "military engineering," "safety engineering / disasters," etc. can be used as information specifying the range of topics. This makes it possible to obtain a report Rpt containing specialized information.

[0082] Furthermore, information specifying the format of the report Rpt can be included in the question Qre. Specifically, "Please report like an expert," "Please report like a researcher," "Please report like an engineer," etc. can be used as information specifying the format of the report Rpt. This allows for the creation of a professionally reasoned report Rpt.

[0083] <<Configuration Example of Component 20>> The component 20 has a function of receiving a directive Pt and transmitting a query Inq and an answer Ans to the component 21. The component 20 also has a function of performing processing using a large-scale language model LLM.

[0084] The large-scale language model LLM has a function of determining whether a question Qre is clear and appropriate according to a directive Pt, and a function of generating an inquiry Inq and an answer Ans.

[0085] For example, Bidirectional Encoder Representations from Transformers (BERT), GPT-3, GPT-3.5, GPT-4 (registered trademark), Language Model for Dialogue Applications (LaMDA), Pathways Language Model (PaLM), Llama2, ALBERT, XLNet, etc. can be used for the large-scale language model LLM.

[0086] <<Configuration Example of Component 21>> The component 21 includes a subcomponent 21A, a subcomponent 21B, and a subcomponent 21C (see FIG. 2).

[0087] [Configuration Example of Subcomponent 21A] The subcomponent 21A includes a search engine SE and a search index Ind. The search engine SE has a function of using a query Qre to find related information RI from the search index Ind, calculating a similarity DoS, and creating a search result list SRL.

[0088] The subcomponent 21A has a function of supplementing and updating the question Qre using the response Res and the reply Rep. In other words, the information processing system according to one aspect of the present invention has a function of supplementing and updating the original question Qre using the response Res to the user and the reply Rep from the user.

[0089] The search result list SRL also includes headings of related information RI sorted in descending order of similarity DoS. The similarity DoS represents the strength of the relationship between the question Qre and the related information RI. For example, a hyperlink can be used to associate the address of the related information RI stored in the archive with a heading in the search result list SRL. This allows the related information RI to be directly viewed from the heading.

[0090] For example, a full-text search system can be used for the subcomponent 21A. The full-text search system can create a search index Ind for documents stored in the archive. The full-text search system can also use the search index Ind to find related information RI related to the question Qre from documents stored in the archive. The full-text search system can also create a search result list SRL. The full-text search system can also handle documents stored in multiple archives across the board.

[0091] Furthermore, documents stored in the archive can be converted into distributed representations (also called embedded representations) in advance using a large-scale language model. The question Qre can also be converted into a distributed representation using a large-scale language model. Using distributed representations makes it possible to handle not only documents written in different languages, but also different types of data as related information. Specifically, image data, audio data, and video data can also be handled.

[0092] Specifically, documents and questions Qre stored in the archive can be converted into distributed representations using BERT, GPT-3, GPT-3.5, GPT-4 (registered trademark), LaMDA, PaLM, Llama2, ALBERT, XLNet, etc.

[0093] By comparing the distributed representation of the document stored in the archive with the distributed representation of the question Qre, semantic similarity can be determined. By evaluating the closeness of the meaning of the document stored in the archive with the meaning of the question Qre, it is possible to absorb fluctuations in expression due to synonyms, etc. In particular, it is possible to absorb fluctuations in the expression of verbs.

[0094] For example, the similarity DoS can be calculated using the Euclidean distance between the distributed representations to be compared. Also, the cosine similarity between the distributed representations to be compared can be used as the similarity DoS.

[0095] The subcomponent 21A also has a function of selecting background information BI from the related information RI based on the similarity DoS. For example, documents from the highest similarity DoS up to a predetermined rank can be selected as background information BI. Specifically, the document with the highest similarity DoS and the document with the second highest similarity DoS can be selected as background information BI.

[0096] For example, an external search service that provides information on websites on the Internet can be used as the search engine SE.

[0097] In addition, archives of public documents, archives of private documents, archives that manage confidential information within the organization to which the user of the information processing system belongs, etc. can be used for document archives. Specifically, patent publications, patent bulletins, books, literature, manuals, confidential documents, and confidential documents can be included in the search targets.

[0098] Product instruction manuals intended for users with advanced technical knowledge require accurate and detailed descriptions. As a result, for example, the user's manual for an electronic design automation (EDA) tool contains a huge amount of text, making even searching difficult for inexperienced users. By including an archive containing such instruction manuals as a search target in an information processing system according to an embodiment of the present invention, even inexperienced users can easily obtain the necessary information. Furthermore, even users who are unable to use technical terminology accurately can easily obtain the necessary information. Furthermore, by adding technical assets accumulated in the user's organization to the archive, not only general usage methods for the product but also specific usage methods used in the organization can be shared with experienced engineers.

[0099] [Configuration Example of Subcomponent 21B] The subcomponent 21B has a function of creating a directive Pt and transmitting it to the component 20.

[0100] The instruction statement Pt includes a question Qre, background information BI, a command g1() for generating an answer Ans to the question Qre by referring to the background information BI, and a command g2() for generating an inquiry statement Inq for generating a more detailed answer Ans, while suspending the generation of the answer Ans as necessary (see FIG. 3 ). Note that the inquiry statement Inq includes a prompt for supplementary information regarding the question Qre and a prompt for confirmation regarding the question Qre. Note that a method of using an instruction statement Pt including the question Qre, background information BI, and a command g1() for generating an answer Ans to the question Qre by referring to the background information BI when having the large-scale language model LLM generate an answer Ans can be considered one aspect of a Retrieval Augmented Generation (RAG) method.

[0101] For example, "Create an answer Ans to the user's question Qre, also referring to the background information BI" can be used in command g1(). Also, for example, "In order to create a more detailed answer Ans to the user's question Qre, if necessary, reserve the creation of the answer Ans and create an inquiry Inq that asks the user for additional information regarding the question Qre or confirmation of the content of the question Qre" can be used in command g2().

[0102] [Configuration Example of Subcomponent 21C] The subcomponent 21C has a function of creating a response Res and a report Rpt and transmitting them to the component 30.

[0103] The response Res includes the query Inq and the search result list SRL (see FIG. 4A ). For example, the search result list SRL can highlight the related information RI used in the background information BI. This can also attract the user's attention. For example, the response Res can include a sentence such as, "Documents related to the question Qre are shown in the search result list SRL below. The question Qre is unclear. Please clarify the question Qre." Furthermore, for example, the response Res can receive similarity DoS from the subcomponent 21A and add the corresponding similarity DoS to each heading of the related information RI in the search result list SRL.

[0104] The report Rpt also includes an answer Ans and a search result list SRL (see FIG. 4B ). For example, the report Rpt includes a sentence such as, "Documents related to the question Qre are shown in the search result list SRL below. Also, with reference to the background information BI, the answer Ans created by the AI ​​is as follows."

[0105] This makes it possible to calculate the similarity DoS even for documents with different attributes. Furthermore, documents with different attributes can be registered in the search index Ind. Furthermore, documents with different attributes can be treated as related information RI across a wide range of documents. Furthermore, because a huge number of documents can be registered in the search index Ind, background information BI can be adopted across a wide range of documents. Furthermore, background information BI can be adopted across a wide range of documents, for example, from official documents, patent publications, patent gazettes, books, literature, manuals, private documents, etc. Furthermore, background information BI can be adopted from confidential documents, confidential documents, etc. Furthermore, background information BI can be adopted from the related information RI based on the similarity DoS, which indicates the strength of the relationship with the question Qre. If the question Qre is unclear, generation of an answer Ans can be reserved. Furthermore, if the question Qre is inappropriate, generation of an answer Ans can be reserved. Furthermore, unnecessary processing using the large-scale language model LLM can be eliminated. Furthermore, the question Qre can be improved to be clear and appropriate through dialogue using the response Res and reply Rep. Furthermore, through dialogue using responses Res and replies Rep, for example, it is possible to supplement the user's linguistic expression ability. Furthermore, it is possible to acquire related information RI using clear and appropriate questions Qre. Furthermore, it is possible to reduce noise contained in the related information RI. As a result, it is possible to provide a novel information processing system that is highly convenient, useful, and reliable.

[0106] <Configuration Example 2 of Information Processing System> In addition, the information processing system according to one aspect of the present invention described in this embodiment has a function in which the component 30 receives a stop command AC and transmits it to the component 21. For example, a user who notices that unintended related information RI is used in the background information BI can input the stop command AC to the component 30.

[0107] The component 21 also has a function of sending an abort command AC to the component 20 .

[0108] The component 20 also has a function of accepting an abort command AC and aborting processing using the large-scale language model LLM.

[0109] This allows the processing using the large-scale language model LLM to be stopped, thereby eliminating wasteful processing. For example, it is possible to eliminate wasteful processing time and power consumption. As a result, it is possible to provide a novel information processing system that is highly convenient, useful, and reliable.

[0110] <Configuration Example 3 of Information Processing System> In addition, in the information processing system according to one aspect of the present invention described in this embodiment, the component 30 has a function of accepting the selection information SI and transmitting it to the component 21. The component 21 also has a function of updating the background information BI using the selection information SI, and the background information BI includes related information RI adopted based on the selection information SI instead of the similarity DoS.

[0111] For example, if a user notices that related information RI that should have been adopted in the background information BI has not been adopted, the user can input selection information SI to the component 30. Alternatively, if a user notices that unintended related information RI has been adopted in the background information BI, the user can input selection information SI to the component 30. Furthermore, the selection information SI can be used to re-adopt or cancel the adoption of the background information BI.

[0112] As a result, even if unintended background information BI is adopted, it can be re-adopted using the selected information SI. Furthermore, even if unintended background information BI is adopted due to, for example, the user's linguistic expression ability, it can be re-adopted using the selected information SI. Furthermore, even if unintended background information BI is adopted due to, for example, the bias of information registered in the search index Ind, it can be re-adopted using the selected information SI. As a result, it is possible to provide a novel information processing system that is excellent in convenience, usefulness, and reliability.

[0113] <Configuration Example 4 of Information Processing System> The information processing system described in this embodiment includes a component 30, a component 20, and a component 21 (see FIG. 1).

[0114] An information processing system according to one embodiment of the present invention includes one or more information processing devices. An information processing device shared by multiple information processing systems can be used for the information processing system according to one embodiment of the present invention. For example, multiple information processing devices connected via a network 51 can be used for the information processing system according to one embodiment of the present invention. Note that when the information processing system according to one embodiment of the present invention is configured using multiple information processing devices, the load related to information processing can be distributed.

[0115] For example, an information processing system according to one aspect of the present invention can be configured with an information processing device that performs the functions of component 30, an information processing device that performs the functions of component 21, and an information processing device that performs the functions of component 20. Note that a plurality of information processing devices can be used to perform the functions of one component.

[0116] Configuration Example 1 of Information Processing Device Configuration Example 1 of the information processing device described in this embodiment can be used for the component 30. Configuration Example 1 of the information processing device can also be called a client computer, etc. For example, a desktop computer can be used for the component 30.

[0117] The information processing device according to the first exemplary configuration can accept data input by a user of the information processing system according to an embodiment of the present invention. The information processing device according to the first exemplary configuration can also provide the user with data output by the information processing system according to an embodiment of the present invention.

[0118] In the first exemplary configuration of the information processing device, for example, dedicated application software or a web browser runs. A user of the information processing system according to an embodiment of the present invention can access the information processing system via either of these. This allows the user to enjoy services using the information processing system according to an embodiment of the present invention.

[0119] Configuration Example 2 of Information Processing Apparatus Configuration Example 2 of the information processing apparatus described in this embodiment can be used for the component 21. For example, the component 21 can be a workstation, a server computer, a supercomputer, or the like.

[0120] The information processing device in configuration example 2 preferably has a function as a parallel computer. By using the information processing device as a parallel computer, it is possible to perform large-scale calculations necessary for learning and inference of artificial intelligence (AI), for example.

[0121] Furthermore, configuration example 2 of the information processing device can perform processing using a natural language processing model that uses AI.

[0122] For example, processing can be performed using natural language processing models such as BERT, T5 (Text-to-Text Transfer Transformer), GPT-3, GPT-3.5, GPT-4 (registered trademark), LaMDA, PaLM, and Llama2.

[0123] Configuration Example 3 of Information Processing Device Configuration Example 3 of the information processing device described in this embodiment can be used for component 20. Note that component 20 is larger in scale and has higher computing power than component 21. For example, a large computer such as a server computer or a supercomputer can be used for component 20.

[0124] The information processing device in configuration example 3 preferably has a function as a parallel computer. By using the information processing device as a parallel computer, it is possible to perform large-scale calculations necessary for AI learning and inference, for example.

[0125] Furthermore, the information processing device configuration example 3 can perform processing using a natural language processing model using AI, particularly processing using a general-purpose language processing model that can perform various natural language processing tasks.

[0126] For example, processing can be performed using natural language models such as BERT, T5, GPT-3, GPT-3.5, GPT-4 (registered trademark), LaMDA, PaLM, and Llama2. In particular, it is preferable to be able to perform processing using GPT-4 (registered trademark). For example, if processing can be performed using a large-scale language model that is larger than conventional natural language models, more natural sentence generation or dialogue can be realized.

[0127] Note that a person who provides a service using an information processing system according to one embodiment of the present invention does not necessarily have to own the information processing device of Configuration Example 3. For example, a service provider can use part of a service provided by another business or the like using the information processing device of Configuration Example 3.

[0128] The network 51 that can be used in the information processing system of one embodiment of the present invention can connect multiple information processing devices. This allows the connected multiple information processing devices to transmit and receive data to and from each other. In addition, the load related to information processing can be distributed.

[0129] When wireless communication is performed, communication standards such as the fourth generation mobile communication system (4G), fifth generation mobile communication system (5G), and sixth generation mobile communication system (6G), or specifications standardized by the IEEE such as Wi-Fi (registered trademark) and Bluetooth (registered trademark), can be used as communication protocols or communication technologies.

[0130] For example, a local network can be used for the network 51. Also, an intranet or an extranet can be used for the network 51. Also, a personal area network (PAN), a local area network (LAN), a campus area network (CAN), a metropolitan area network (MAN), a wide area network (WAN), a global area network (GAN), etc. can be used for the network 51.

[0131] Furthermore, for example, a global network can be used for the network 51. Specifically, the Internet, which is the foundation of the World Wide Web (WWW), can be used.

[0132] Furthermore, a person who provides a service using the information processing system according to one embodiment of the present invention can provide the service using the information processing method according to one embodiment of the present invention via the network 51, for example.

[0133] When the information processing system according to an embodiment of the present invention is built within a local network, the possibility of confidential information leaking can be reduced, for example, compared to when the Internet is used.

[0134] <<Configuration Example 4 of Information Processing Device>> An information processing device that can be used in an information processing system of one embodiment of the present invention includes, for example, an input unit 110, a memory unit 120, a processing unit 130, an output unit 140, and a transmission path 150 (see FIG. 5).

[0135] In the drawings accompanying this specification, the components are classified by function and shown as independent blocks in the block diagrams. However, in reality, it is difficult to completely separate the components by function, and one component may be involved in multiple functions. For example, part of the processing unit 130 may function as the input unit 110. Also, one function may be involved in multiple components. For example, the processing performed by the processing unit 130 may be executed by different servers depending on the processing.

[0136] [Input Unit 110] The input unit 110 can receive data from outside the information processing device. For example, the input unit 110 receives data via the network 51. Specifically, a device such as a personal computer equipped with a communication port or a communication function can be used.

[0137] The input unit 110 supplies the received data to one or both of the storage unit 120 and the processing unit 130 via the transmission path 150 .

[0138] [Storage Unit 120] The storage unit 120 has a function of storing a program executed by the processing unit 130. The storage unit 120 can also have a function of storing data generated by the processing unit 130 (e.g., calculation results, analysis results, inference results), data accepted by the input unit 110, etc.

[0139] The storage unit 120 may have a database. Furthermore, the information processing device may have a database separate from the storage unit 120. The information processing device may have a function to retrieve data from a database that exists outside the storage unit 120, outside the information processing device, or outside the information processing system. Furthermore, the information processing device may have a function to retrieve data from both its own database and an external database.

[0140] Either or both of a storage and a file server can be used as the memory unit 120. Also, a database that records paths of files stored in a file server can be used as the memory unit 120.

[0141] The storage unit 120 includes at least one of a volatile memory and a non-volatile memory. Examples of the volatile memory include a dynamic random access memory (DRAM) and a static random access memory (SRAM). Examples of the non-volatile memory include a resistive random access memory (ReRAM), a phase change random access memory (PRAM), a ferroelectric random access memory (FeRAM), a magnetoresistive random access memory (MRAM), and a flash memory. The storage unit 120 may include at least one of NOSRAM (registered trademark) and DOSRAM (registered trademark). The storage unit 120 may include a recording media drive. Examples of the recording media drive include a hard disk drive (HDD) and a solid state drive (SSD).

[0142] NOSRAM is an abbreviation for "Nonvolatile Oxide Semiconductor Random Access Memory (RAM)." NOSRAM refers to a memory in which memory cells are two-transistor (2T) or three-transistor (3T) gain cells and transistors (also called OS transistors) that use metal oxide in their channel formation regions. OS transistors have an extremely small leakage current, i.e., a current that flows between the source and drain in an off state. NOSRAM can be used as a nonvolatile memory by retaining a charge corresponding to data in the memory cell using its extremely small leakage current characteristic. In particular, NOSRAM can read stored data without destroying it (nondestructive readout), making it suitable for arithmetic processing in which only data read operations are repeated a large number of times. NOSRAM can increase its data capacity by stacking layers, and therefore can be used as a large-scale cache memory, main memory, or storage memory to improve the performance of semiconductor devices.

[0143] DOSRAM is an abbreviation for "Dynamic Oxide Semiconductor RAM" and refers to a RAM having 1T (transistor) 1C (capacitor) type memory cells. DOSRAM is a DRAM formed using OS transistors, and is a memory that temporarily stores information sent from an external device. DOSRAM is a memory that takes advantage of the low off-state current of OS transistors.

[0144] In this specification and the like, a metal oxide refers to an oxide of a metal in a broad sense. Metal oxides are classified into oxide insulators, oxide conductors (including transparent oxide conductors), oxide semiconductors (also referred to as oxide semiconductors or simply as OSs), and the like. For example, when a metal oxide is used for a semiconductor layer of a transistor, the metal oxide may be referred to as an oxide semiconductor.

[0145] The metal oxide included in the channel formation region preferably contains indium (In). When the metal oxide included in the channel formation region contains indium, the carrier mobility (electron mobility) of the OS transistor is increased. Furthermore, the metal oxide included in the channel formation region is preferably an oxide semiconductor containing element M. The element M is preferably at least one of aluminum (Al), gallium (Ga), and tin (Sn). Other elements applicable to element M include boron (B), silicon (Si), titanium (Ti), iron (Fe), nickel (Ni), germanium (Ge), yttrium (Y), zirconium (Zr), molybdenum (Mo), lanthanum (La), cerium (Ce), neodymium (Nd), hafnium (Hf), tantalum (Ta), and tungsten (W). However, a combination of two or more of the above elements may be used as element M. The element M is, for example, an element having a high bond energy with oxygen. For example, it is an element having a higher bond energy with oxygen than indium. The metal oxide contained in the channel formation region is preferably a metal oxide containing zinc (Zn), since zinc-containing metal oxides may be easily crystallized.

[0146] The metal oxide contained in the channel formation region is not limited to a metal oxide containing indium, but may be, for example, a metal oxide containing zinc but not indium, such as zinc tin oxide or gallium tin oxide, a metal oxide containing gallium, or a metal oxide containing tin.

[0147] [Processing Unit 130] The processing unit 130 has a function of performing processes such as calculation, analysis, and inference using data supplied from one or both of the input unit 110 and the storage unit 120. The processing unit 130 can supply generated data (e.g., calculation results, analysis results, and inference results) to one or both of the storage unit 120 and the output unit 140.

[0148] The processing unit 130 has a function of acquiring data from the storage unit 120. The processing unit 130 can also have a function of recording or registering data in the storage unit 120.

[0149] The processing unit 130 may include, for example, an arithmetic circuit. The processing unit 130 may include, for example, a central processing unit (CPU). The processing unit 130 may also include a graphics processing unit (GPU). The processing unit 130 may also include a neural processing unit / neural network processing unit (NPU).

[0150] The processing unit 130 may include a microprocessor such as a DSP (Digital Signal Processor). The microprocessor may be implemented by a PLD (Programmable Logic Device) such as an FPGA (Field Programmable Gate Array) or an FPAA (Field Programmable Analog Array). The processing unit 130 may also include a quantum processor. The processing unit 130 can perform various data processing and program control by interpreting and executing instructions from various programs using the processor. Programs that can be executed by the processor are stored in at least one of the memory area of ​​the processor and the storage unit 120.

[0151] The processing unit 130 may include a main memory. The main memory may include at least one of a volatile memory such as a RAM and a non-volatile memory such as a ROM (Read Only Memory). The main memory may also include at least one of the above-mentioned NOSRAM and DOSRAM.

[0152] The RAM may be, for example, a DRAM or an SRAM, and a virtual memory space is allocated and used as a working space for the processing unit 130. The operating system, application programs, program modules, program data, lookup tables, and the like stored in the storage unit 120 are loaded into the RAM for execution. The data, programs, and program modules loaded into the RAM are each directly accessed and operated by the processing unit 130.

[0153] The ROM can store a BIOS (Basic Input / Output System), firmware, etc., which do not require rewriting. Examples of ROM include mask ROM, OTPROM (One Time Programmable Read Only Memory), and EPROM (Erasable Programmable Read Only Memory). Examples of EPROMs include UV-EPROMs (Ultra-Violet Erasable Programmable Read Only Memories), which allow stored data to be erased by exposure to ultraviolet light, EEPROMs (Electrically Erasable Programmable Read Only Memories), and flash memories.

[0154] The processing section 130 can include one or both of an OS transistor and a transistor having silicon in a channel formation region (a Si transistor).

[0155] The processing unit 130 preferably includes an OS transistor. Because an OS transistor has an extremely small off-state current, using the OS transistor as a switch for retaining charge (data) flowing into a capacitor functioning as a memory element can ensure a long data retention period. By using this characteristic in at least one of the register and cache memory of the processing unit, the processing unit can be operated only when necessary, and can be turned off in other cases by saving information from the previous processing in the memory element. In other words, normally-off computing is possible, and the power consumption of the information processing system can be reduced.

[0156] It is preferable that the information processing device uses AI for at least some of its processing.

[0157] It is particularly preferable that the information processing device uses an artificial neural network (ANN, hereinafter simply referred to as a neural network). A neural network is realized by a circuit (hardware) or a program (software).

[0158] In this specification, a neural network refers to a general model that mimics the neural circuit network of a living organism, determines the connection strength between neurons through learning, and has problem-solving capabilities. A neural network has an input layer, an intermediate layer (hidden layer), and an output layer.

[0159] In this specification and the like, when discussing neural networks, determining the connection strengths (also called weighting coefficients) between neurons from existing information may be referred to as "learning."

[0160] In this specification and the like, the act of constructing a neural network using connection strengths obtained by learning and deriving a new conclusion from it may be referred to as "inference."

[0161] [Output Unit 140] The output unit 140 can output at least one of the calculation results, analysis results, and inference results from the processing unit 130 to the outside of the information processing device. For example, the output unit 140 can transmit data via the network 51. Specifically, a device such as a personal computer equipped with a communication port or a communication function can be used. Furthermore, a device equipped with a communication function may be used for the input unit 110 and the output unit 140.

[0162] [Transmission Path 150] The transmission path 150 has a function of transmitting data. Data can be transmitted and received between the input unit 110, the storage unit 120, the processing unit 130, and the output unit 140 via the transmission path 150. Specifically, an external bus, a LAN, or the Internet can be used as the transmission path 150.

[0163] Note that this embodiment mode can be appropriately combined with other embodiment modes described in this specification.

[0164] Embodiment 2 In this embodiment, an information processing method according to one embodiment of the present invention will be described with reference to FIGS. 6 and 7. FIG.

[0165] FIG. 6 is a diagram illustrating an information processing method according to one aspect of the present invention.

[0166] FIG. 7 is a diagram illustrating an information processing method according to one embodiment of the present invention, which is different from the information processing method described with reference to FIG.

[0167] <Example 1 of Information Processing Method> An information processing method according to one embodiment of the present invention includes steps S1 to S14 (see FIG. 6).

[0168] [Step S1] In step S1, the component 30 accepts a question Qre from a user and transmits it to the component 21.

[0169] [Step S2] In step S2, the component 21 receives the question Qre and creates a search result list SRL.

[0170] The component 21 includes a search engine SE and a search index Ind. The search engine SE has a function of using a query Qre to find related information RI from the search index Ind and calculating a similarity DoS.

[0171] The search result list SRL includes the headings of the related information RI sorted in descending order of similarity DoS, which represents the strength of the relationship between the question Qre and the related information RI.

[0172] [Step S3] In step S3, the component 21 adopts the background information BI from the related information RI based on the similarity DoS.

[0173] [Step S4] In step S4, the component 21 creates a response Res and passes it to the component 30. The response Res includes the search result list SRL.

[0174] [Step S5] In step S5, the component 30 provides a response Res to the user.

[0175] [Step S6] In step S6, the component 21 creates an instruction Pt and transmits it to the component 20. The instruction Pt includes a question Qre, background information BI, and a command g1( ) for generating an answer Ans to the question Qre by referring to the background information BI.

[0176] [Step S7] In step S7, the component 20 receives the directive Pt and starts generating an answer Ans using the large-scale language model LLM.

[0177] The component 20 has a function of performing processing using a large-scale language model LLM. The large-scale language model LLM has a function of generating an answer Ans according to a directive Pt.

[0178] [Step S8] In step S8, if component 30 receives selection information SI from the user before component 20 completes generation of answer Ans, component 30 passes the selection information SI to component 21 and proceeds to step S9. Otherwise, the process proceeds to step S12.

[0179] [Step S9] In step S9, the component 21 passes the stop command AC to the component 20, and re-adopts the background information BI from the related information RI based on the selection information SI.

[0180] [Step S10] In step S10, the component 20 receives the stop command AC and stops generating the answer Ans.

[0181] [Step S11] In step S11, the component 21 updates the response Res, passes it to the component 30, and proceeds to step S5. The updated response Res includes the search result list SRL and the selection information SI.

[0182] [Step S12] In step S12, the component 20 transmits the answer Ans to the component 21.

[0183] [Step S13] In step S13, the component 21 receives the answer Ans, creates a report Rpt, and transmits it to the component 30. The report Rpt includes the answer Ans and the search result list SRL.

[0184] [Step S14] In step S14, the component 30 provides the report Rpt to the user.

[0185] This makes it possible to calculate the similarity DoS even for documents with different attributes. Furthermore, documents with different attributes can be registered in the search index Ind. Furthermore, documents with different attributes can be treated as related information RI across a wide range of documents. Furthermore, because a huge number of documents can be registered in the search index Ind, background information BI can be adopted across a wide range of documents. Furthermore, background information BI can be adopted across a wide range of documents, for example, from official documents, patent publications, patent gazettes, books, literature, manuals, private documents, etc. Furthermore, background information BI can be adopted from confidential documents, confidential documents, etc. Furthermore, background information BI can be adopted from the related information RI based on the similarity DoS, which indicates the strength of the relationship with the question Qre. Furthermore, even if unintended background information BI is adopted, it can be re-adopted using the selected information SI. Furthermore, even if unintended background information BI is adopted due to, for example, the user's linguistic expression ability, it can be re-adopted using the selected information SI. Furthermore, even if unintended background information BI is adopted due to a bias in the information registered in the search index Ind, for example, it can be re-adopted using the selected information SI. Furthermore, the large-scale language model LLM can generate an answer Ans while examining whether the adopted background information BI is appropriate. Furthermore, time is not wasted examining whether the adopted background information BI is appropriate. As a result, a novel information processing method that is highly convenient, useful, and reliable can be provided.

[0186] If the component 30 receives selection information SI from the user after the component 20 has generated the answer Ans, the component 30 can pass the selection information SI to the component 21 and proceed to step S9. This allows the user to re-adopt the background information BI after reviewing the report Rpt, for example.

[0187] <Example 2 of Information Processing Method> An information processing method according to one embodiment of the present invention includes steps S1 to S21 (see FIG. 7).

[0188] [Step S1] In step S1, the component 30 accepts a question Qre from a user and transmits it to the component 21.

[0189] [Step S2] In step S2, the component 21 receives the question Qre and creates a search result list SRL.

[0190] The component 21 includes a search engine SE and a search index Ind. The search engine SE has a function of using a query Qre to find related information RI from the search index Ind and calculating a similarity DoS.

[0191] The search result list SRL includes the headings of the related information RI sorted in descending order of similarity DoS, which represents the strength of the relationship between the question Qre and the related information RI.

[0192] [Step S3] In step S3, the component 21 adopts the background information BI from the related information RI based on the similarity DoS.

[0193] [Step S4] In step S4, the component 21 creates a response Res and passes it to the component 30. The response Res includes the search result list SRL.

[0194] [Step S5] In step S5, the component 30 provides a response Res to the user.

[0195] [Step S6] In step S6, the component 21 creates an instruction Pt and sends it to the component 20. The instruction Pt includes a question Qre, background information BI, an instruction g1() for generating an answer Ans to the question Qre by referring to the background information BI, and an instruction g2() for generating an inquiry Inq for generating a more detailed answer Ans, suspending the generation of the answer Ans as necessary. The inquiry Inq also includes a prompt for supplementary information regarding the question Qre and a prompt for confirmation regarding the question Qre.

[0196] [Step S7] In step S7, the component 20 receives the directive Pt and considers the question Qre using the large-scale language model LLM.

[0197] The component 20 has a function of performing processing using a large-scale language model LLM, which has a function of determining whether a question Qre is clear and appropriate according to a directive Pt, and a function of generating a query Inq and an answer Ans.

[0198] [Step S8] If the large scale language model LLM determines in step S8 that the question Qre is unclear or inappropriate, the process proceeds to step S9. If the large scale language model LLM determines that the question Qre is clear and appropriate, the process proceeds to step S14.

[0199] [Step S9] In step S9, the component 20 generates an inquiry statement Inq and sends it to the component 21.

[0200] [Step S10] In step S10, the component 21 updates the response Res and sends it to the component 30. The updated response Res includes the query Inq and the search result list SRL.

[0201] [Step S11] In step S11, the component 30 provides a response Res to the user.

[0202] [Step S12] In step S12, the component 30 receives a reply Rep from the user and transmits it to the component 21.

[0203] [Step S13] In step S13, the component 21 receives the response Res, updates the question Qre, and proceeds to step S2. The updated question Qre includes the response Res and the reply Rep.

[0204] [Step S14] In step S14, the component 20 starts generating an answer Ans using the large-scale language model LLM.

[0205] [Step S15] In step S15, if component 30 receives selection information SI from the user before component 20 completes generation of answer Ans, component 30 passes the selection information SI to component 21 and proceeds to step S16. Otherwise, the process proceeds to step S19.

[0206] [Step S16] In step S16, the component 21 passes the stop command AC to the component 20, and re-adopts the background information BI from the related information RI based on the selection information SI.

[0207] [Step S17] In step S17, the component 20 receives the stop command AC and stops generating the answer Ans.

[0208] [Step S18] In step S18, the component 21 updates the response Res, passes it to the component 30, and proceeds to step S5. The updated response Res includes the search result list SRL and the selection information SI.

[0209] [Step S19] In step S19, the component 20 transmits the answer Ans to the component 21.

[0210] [Step S20] In step S20, the component 21 receives the answer Ans, creates a report Rpt, and transmits it to the component 30. The report Rpt includes the answer Ans and the search result list SRL.

[0211] [Step S21] In step S21, the component 30 provides a report Rpt.

[0212] As a result, generation of an answer Ans can be withheld when the question Qre is unclear. Furthermore, generation of an answer Ans can be withheld when the question Qre is inappropriate. Furthermore, unnecessary processing using the large-scale language model LLM can be avoided. Furthermore, through a dialogue using the response Res and reply Rep, the question Qre can be improved to be clear and appropriate. Furthermore, through a dialogue using the response Res and reply Rep, for example, the user's linguistic expression ability can be supplemented. Furthermore, related information RI can be obtained using a clear and appropriate question Qre. Furthermore, noise contained in the related information RI can be reduced. As a result, a novel information processing method that is highly convenient, useful, and reliable can be provided.

[0213] If the component 30 receives selection information SI from the user after the component 20 has generated the answer Ans, the component 30 can pass the selection information SI to the component 21 and proceed to step S16. This allows the user to re-adopt the background information BI after reviewing the report Rpt, for example.

[0214] Note that this embodiment mode can be appropriately combined with other embodiment modes described in this specification.

[0215] AC: abort command, Ans: answer, BI: background information, Ind: search index, Inq: query, LLM: large scale language model, Pt: instruction, Qre: question, Rep: reply, Res: response, Rpt: report, SE: search engine, SI: selection information, SRL: search result list, 20: component, 21: component, 21A: subcomponent, 21B: subcomponent, 21C: subcomponent, 30: component, 51: network, 110: input unit, 120: memory unit, 130: processing unit, 140: output unit, 150: transmission path

Claims

a first component, a second component, and a third component, and the first component is provided with a function of receiving questions and replies and transmitting them to the third component, the questions and the replies are described in natural language, the first component is provided with a function of receiving responses and reports and providing them, the second component is provided with a function of receiving instructions and transmitting inquiry sentences and answers to the third component, the second component is provided with a function of performing processing using a large language model, the large language model is provided with a function of determining whether the question is clear and appropriate according to the instruction, and a function of generating the inquiry sentence and the answer, the third component includes a first sub-component, a second sub-component, and a third sub-component, the first sub-component is provided with a function of supplementing and updating the question using the response and the reply, a search engine, and a search index, the search engine is provided with a function of using the question to find relevant information from the search index, calculating a similarity, and creating a list of search results, the list of search results includes headings of the relevant information arranged in descending order of the similarity, the similarity represents the strength of the relationship between the question and the relevant information, the first sub-component is provided with a function of adopting background information from the relevant information based on the similarity, the second sub-component is provided with a function of creating the instruction and transmitting it to the second component, the instruction includes the question, the background information, a first instruction for generating an answer to the question with reference to the background information, and a second instruction for generating the inquiry sentence for generating a more detailed answer by withholding the generation of the answer if necessary, the inquiry sentence includes a request for supplementation related to the question and a request for confirmation related to the question, the third sub-component is provided with a function of creating the response and the report and transmitting them to the first component, the response includes the inquiry sentence and the list of search results, the report includes the answer and the list of search results, an information processing system. 　 The first component is configured to receive a stop command and transmit it to the third component. The third component is configured to transmit the stop command to the second component. The information processing system according to claim 1, wherein the second component is configured to receive the stop command and stop the processing using the large language model. 　 The first component is configured to receive selection information and transmit it to the third component. The information processing system according to claim 1, wherein the background information includes the relevant information adopted based on the selection information instead of the similarity. 　 An information processing method having steps from the first step to the fourteenth step, In the first step, a first component receives a question and transfers it to a second component. In the second step, the second component receives the question and creates a list of search results. The second component includes a search engine and a search index. The search engine is configured to use the question to find relevant information from the search index and calculate a similarity. The list of search results includes headings of the relevant information arranged in descending order of the similarity. The similarity represents the strength of the relationship between the question and the relevant information. In the third step, the second component adopts background information from the relevant information based on the similarity. In the fourth step, the second component creates a response and transfers it to the first component. The response includes the list of search results. In the fifth step, the first component provides the response. In the sixth step, the second component creates an instruction text and transfers it to a third component. The instruction text includes the question, the background information, and an instruction to generate an answer to the question with reference to the background information. In the seventh step, the third component receives the instruction text and starts generating the answer using a large language model. The third component is configured to perform processing using the large language model. The large language model is configured to generate the answer according to the instruction text. In the eighth step, if the first component receives selection information before the third component completes the generation of the answer, the first component transfers the selection information to the second component and proceeds to the ninth step of the process; otherwise, it proceeds to the twelfth step of the process. In the ninth step, the second component transfers a cancellation instruction to the third component and re-adopts the background information from the relevant information based on the selection information. In the tenth step, the third component receives the cancellation instruction and cancels the generation of the answer. In the eleventh step, the second component updates the response and transfers it to the first component, proceeding to the fifth step of the process. The updated response includes the search result list and the selection information. In the twelfth step, the third component transfers the answer to the second component. In the thirteenth step, the second component receives the answer, creates a report, and transfers it to the first component. The report includes the answer and the search result list. In the fourteenth step, the first component provides the report, an information processing method. 　 An information processing method having steps from the first step to the twenty-first step, In the first step, a first component receives a question and transfers it to a second component. In the second step, the second component receives the question and creates a search result list. The second component includes a search engine and a search index. The search engine has a function of using the question to find relevant information from the search index and calculating a similarity. The search result list includes headings of the relevant information arranged in descending order of the similarity. The similarity represents the strength of the relationship between the question and the relevant information. In the third step, the second component adopts background information from the relevant information based on the similarity. In the fourth step, the second component creates a response and transfers it to the first component. The response includes the search result list. In the fifth step, the first component provides the response, In the sixth step, the second component creates an instruction and transfers it to a third component, The instruction includes the question, the background information, a first instruction to generate an answer to the question with reference to the background information, and a second instruction to hold off on generating an answer and generate an inquiry statement for generating a more detailed answer if necessary, The inquiry statement includes a supplementary urging related to the question and a confirmation urging related to the question, In the seventh step, the third component receives the instruction and examines the question using a large language model, The third component has a function of performing processing using the large language model, The large language model has a function of determining whether the question is clear and appropriate according to the instruction, and a function of generating the inquiry statement and the answer, In the eighth step, if the large language model determines that the question is not clear or not appropriate, the process proceeds to the ninth step, If the large language model determines that the question is clear and appropriate, the process proceeds to the fourteenth step, In the ninth step, the third component generates the inquiry statement and transfers it to the second component, In the tenth step, the second component updates the response and transfers it to the first component, The response in the tenth step includes the inquiry statement and the list of search results, In the eleventh step, the first component provides the response, In the twelfth step, the first component receives a reply and transfers it to the second component, In the thirteenth step, the second component receives the response, updates the question, and proceeds to the second step, The updated question includes the response and the reply, In the fourteenth step, the third component starts generating the answer using the large language model, In the 15th step, if the first component receives selection information before the third component completes the generation of the answer, the first component passes the selection information to the second component and advances the process to the 16th step; otherwise, the process advances to the 19th step. In the 16th step, the second component passes a stop command to the third component and re - adopts the background information from the related information based on the selection information. In the 17th step, the third component receives the stop command and aborts the generation of the answer. In the 18th step, the second component updates the response, passes it to the first component, and advances the process to the 5th step. The response in the 18th step includes the search result list and the selection information. In the 19th step, the third component passes the answer to the second component. In the 20th step, the second component receives the answer, creates a report, and passes it to the first component. The report includes the answer and the search result list. In the 21st step, the first component provides the report, an information processing method.

Images