Interaction information generation method, system, and electronic device

Abstract

The present disclosure relates to the fields of large model technology and machine learning, and disclosed are an interaction information generation method, a system, and an electronic device. The method is applied to a generative question-answering system, and comprises: inputting query information in an operation interface of the generative question-answering system; using a generative search model to analyze the query information, and generating initial answer content matching the query information; by means of query, obtaining retrieval information associated with the query information, wherein the retrieval information is used for at least representing a result retrieved on the basis of the query information; acquiring associated answer content at least associated with the initial answer content and the retrieval information, wherein the associated answer content is generated by using at least one auxiliary generative search model of the generative search model and the generative search model to execute at least one round of interactive answer; and replacing the initial answer content with the associated answer content to obtain target answer content matching the query information.

Description

Method, system and electronic device for generating interactive information

[0001] Cross-reference

[0002] This disclosure claims priority to a Chinese patent application filed with the Patent Office of China on January 2, 2024, with application number 202410010835.9 and invention name “Method, system and electronic device for generating interactive information”, the entire contents of which are incorporated by reference into this disclosure. Technical Field

[0003] The present disclosure relates to large model technology and machine learning fields, and more specifically, to a method, system, and electronic device for generating interactive information. Background Art

[0004] Currently, generative question-answering systems typically use the search information to determine the response to complex queries. However, because the search information may contain errors, misleading information and hallucinatory responses can occur, limiting search results and leading to low accuracy in the interactive information generated by generative question-answering systems.

[0005] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0006] The embodiments of the present disclosure provide a method, system, and electronic device for generating interaction information, so as to at least solve the technical problem of low accuracy of the generated interaction information.

[0007] According to one aspect of an embodiment of the present disclosure, a method for generating interactive information is provided. The method can be applied to a generative question-answering system and may include: inputting query information into an operation interface of the generative question-answering system; analyzing the query information using a generative search model to generate initial response content that matches the query information; querying to obtain retrieval information associated with the query information, wherein the retrieval information is used to at least represent results retrieved based on the query information; obtaining associated response content associated with at least the initial response content and the retrieval information, wherein the associated response content is generated by executing at least one round of interactive responses using at least one auxiliary generative search model of the generative search model and the generative search model; and replacing the initial response content with the associated response content to obtain target response content that matches the query information.

[0008] According to another aspect of the embodiment of the present disclosure, another method for generating interactive information is also provided. The method can be applied to a generative question-answering system deployed in a search engine, and may include: displaying query information on an operation interface of the generative question-answering system; in response to an initial search instruction acting on the operation interface, displaying initial reply content matching the query information on the operation interface, wherein the initial reply content is obtained by analyzing the query information using a generative search model; in response to a second search instruction acting on the operation interface, displaying target reply content matching the query information on the operation interface, wherein the target reply content is obtained by replacing the initial reply content with associated reply content, the retrieval information is used to at least characterize the results retrieved based on the query information, and the associated reply content is generated by performing at least one round of interactive reply using at least one auxiliary generative search model of the generative search model and the generative search model, and is associated with the initial reply content and the retrieval information.

[0009] According to another aspect of the embodiment of the present disclosure, another method for generating interactive information is also provided. The method may include: inputting query information by calling a first interface, wherein the first interface includes a first parameter, and the parameter value of the first parameter is the query information; analyzing the query information using a generative search model to generate initial reply content that matches the query information; querying to obtain retrieval information associated with the query information, wherein the retrieval information is used to at least characterize the results retrieved based on the query information; obtaining associated reply content associated with at least the initial reply content and the retrieval information, wherein the associated reply content is generated by performing at least one round of interactive reply using at least one auxiliary generative search model and a generative search model of the generative search model; replacing the initial reply content with the associated reply content to obtain the target reply content that matches the query information; outputting the target reply content by calling a second interface, wherein the second interface includes a second parameter, and the parameter value of the second parameter is the target reply content.

[0010] According to another aspect of an embodiment of the present disclosure, a system for generating interactive information is also provided. The system may include: a first information generation end, configured to obtain query information input into an operation interface of a generative question-answering system; analyze the query information using a generative search model to generate initial response content that matches the query information; a second information generation end, configured to perform at least one round of interactive responses using at least one auxiliary generative search model of the generative search model and the generative search model to generate associated response content that is at least associated with the initial response content and retrieval information of the query information, wherein the retrieval information is used to at least represent a result retrieved based on the query information; wherein the first information generation end is configured to replace the initial response content with the associated response content and output target response content that matches the query information.

[0011] According to another aspect of an embodiment of the present disclosure, an electronic device is also provided, which may include a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions. When the above-mentioned computer-executable instructions are executed by the processor, the above-mentioned method of any one of the above-mentioned items is implemented.

[0012] According to another aspect of an embodiment of the present disclosure, a processor is further provided, which is used to run a program, wherein any one of the above methods is executed when the program is running.

[0013] According to another aspect of an embodiment of the present disclosure, a computer-readable storage medium is further provided, the computer-readable storage medium including a stored program, wherein when the program is executed, the device where the storage medium is located is controlled to execute any one of the above methods.

[0014] In an embodiment of the present disclosure, a query is input into an operation interface of a generative question-answering system; the query is analyzed using a generative search model to generate initial response content that matches the query; retrieval information associated with the query is obtained through querying, wherein the retrieval information is used to at least represent the results retrieved based on the query; associated response content associated with at least the initial response content and the retrieval information is obtained, wherein the associated response content is generated by executing at least one round of interactive responses using at least one auxiliary generative search model of the generative search model and the generative search model; and the initial response content is replaced with the associated response content to obtain target response content that matches the query. That is, in this embodiment, an auxiliary generative search model is proposed, wherein the generative search model is first used to analyze the query to obtain the initial response content, and then the retrieval information associated with the query is obtained through querying, and the associated response content generated by the interactive responses of the auxiliary generative search model and the generative search model is replaced with the initial response content to obtain the final target response content, thereby achieving the technical effect of improving the accuracy of the generated interactive information and solving the technical problem of low accuracy of the generated interactive information.

[0015] It is easy to note that the above general description and the following detailed description are only for the purpose of exemplifying and explaining the present disclosure, and do not constitute a limitation of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present disclosure and constitute a part of the present disclosure. The exemplary embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation of the present disclosure. In the drawings:

[0017] FIG1 is a schematic diagram of an application scenario of a method for generating interaction information according to an embodiment of the present disclosure;

[0018] FIG2 is a block diagram of a computing environment according to an embodiment of the present disclosure;

[0019] FIG3 is a flow chart of a method for generating interaction information according to an embodiment of the present disclosure;

[0020] FIG4 is a flowchart of another embodiment of the present disclosure according to the generation of interactive information;

[0021] FIG5 is a flowchart of another method for generating interaction information according to an embodiment of the present disclosure;

[0022] FIG6 is a schematic diagram of a generative question answering system according to an embodiment of the present disclosure;

[0023] FIG7 is a flow chart of a large model generative search method based on an agent community according to an embodiment of the present disclosure;

[0024] FIG8 is a hardware structure block diagram of a computer terminal (or mobile device) according to a method for generating interaction information according to an embodiment of the present disclosure;

[0025] FIG9 is a schematic diagram of a device for generating interaction information according to an embodiment of the present disclosure;

[0026] FIG10 is a schematic diagram of another device for generating interaction information according to an embodiment of the present disclosure;

[0027] FIG11 is a schematic diagram of another device for generating interaction information according to an embodiment of the present disclosure;

[0028] FIG12 is a structural block diagram of a computer terminal according to an embodiment of the present disclosure;

[0029] FIG13 is a block diagram of an electronic device according to a method for generating interaction information according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0030] In order to enable those skilled in the art to better understand the solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the embodiments described are only part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present disclosure.

[0031] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way are interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or components is not necessarily limited to those steps or components clearly listed, but may include other steps or components that are not clearly listed or inherent to these processes, methods, products or devices.

[0032] The technical solutions provided by this disclosure can be implemented using large-scale model technology. Large models, also known as foundation models, are pre-trained on large-scale unlabeled corpora to produce pre-trained models with more than 100 million parameters. Such models can adapt to a wide range of downstream tasks and have good generalization capabilities, such as large-scale language models (LLMs) and multi-modal pre-training models.

[0033] It should be noted that when the large model is actually applied, the pre-trained model can be fine-tuned through a small number of samples, so that the large model can be applied to different tasks. For example, the large model can be widely used in natural language processing (NLP), computer vision, speech processing and other fields. Specifically, it can be applied to computer vision tasks such as visual question answering (VQA), image caption (IC), image generation, etc. It can also be widely used in natural language processing tasks such as text-based sentiment classification, text summary generation, and machine translation. Therefore, the main application scenarios of the large model include but are not limited to digital assistants, intelligent robots, search, online education, office software, e-commerce, intelligent design, etc. In the embodiment of the present disclosure, data processing through a machine learning model in a dialogue scenario is used as an example for explanation.

[0034] First, some nouns or terms that appear in the description of the embodiments of the present disclosure are subject to the following explanations:

[0035] Large models can be large-scale pre-trained models based on deep learning. Large models can be deep learning models with large-scale model parameters, which can include hundreds of millions, tens of billions, hundreds of billions, trillions, or even more than ten trillion model parameters.

[0036] Generative search can refer to the use of large models combined with search information to generate answers;

[0037] An agent community, which can include at least two agent roles, can be used to collaboratively generate accurate search results and answers;

[0038] Retrieval information can be related information retrieved based on the user's query information (query), which can be used to guide the large model to generate a response;

[0039] Hallucination answers can be wrong answers generated by the large model, which are answers contrary to the facts.

[0040] According to a method of an embodiment of the present disclosure, a method for generating interactive information is provided. As an optional implementation, the above-mentioned method for generating interactive information may include but is not limited to being applied to an application scenario as shown in Figure 1. Figure 1 is a schematic diagram of an application scenario of a method for generating interactive information according to an embodiment of the present disclosure. As shown in Figure 1, in the application scenario, the terminal device 12 may, but is not limited to, communicate with the server 16 through the network 14. For example, it can be used to transmit query information, target response content, etc. The server 16 may, but is not limited to, perform operations on the database 18, such as write data operations or read data operations. The above-mentioned terminal device 12 may, but is not limited to, include a human-computer interaction screen, a processor, and a memory. The above-mentioned human-computer interaction screen may, but is not limited to, be used to display query information and target response content on the terminal device 12. The above-mentioned processor may include, but is not limited to, being used to respond to the above-mentioned human-computer interaction operation, perform corresponding operations, or generate corresponding instructions and send the generated instructions to the server 16. The above-mentioned memory is used to store relevant processing data, such as retrieval information, etc.

[0041] As an optional method, the following steps in the method for generating interactive information can be executed on the server 16: Step S102, input query information in the operation interface of the generative question-answering system; Step S104, analyze the query information using the generative search model to generate initial reply content that matches the query information; Step S106, query to obtain retrieval information associated with the query information, wherein the retrieval information is used to at least characterize the results retrieved based on the query information; Step S108, obtain associated reply content that is at least associated with the initial reply content and the retrieval information, wherein the associated reply content is generated by performing at least one round of interactive reply using at least one auxiliary generative search model and a generative search model of the generative search model; Step S110, replace the initial reply content with the associated reply content to obtain the target reply content that matches the query information.

[0042] Using the above method, an auxiliary generative search model is proposed. The generative search model is used to first analyze the query information. After obtaining the initial reply content, the query is obtained by querying the retrieval information associated with the query information. The associated reply content generated by the interactive reply between the auxiliary generative search model and the generative search model is used to replace the initial reply content to obtain the final target reply content, thereby achieving the technical effect of improving the accuracy of the generated interactive information and solving the technical problem of low accuracy of the generated interactive information.

[0043] In another optional embodiment, FIG2 shows in a block diagram an embodiment of using a computer terminal (or mobile device) as a computing node in a computing environment 201. FIG2 is a structural block diagram of a computing environment according to an embodiment of the present disclosure. As shown in FIG2 , the computing environment 201 includes multiple (shown in the figure as 210-1, 210-2, ...) computing nodes (such as servers) running on a distributed network. The computing nodes all contain local processing and memory resources, and the end user 202 can remotely run applications or store data in the computing environment 201. The application can be provided as multiple services 220-1, 220-2, 220-3 and 220-4 in the computing environment 201, representing services "A", "D", "E" and "H" respectively.

[0044] End user 202 can provide and access services through a web browser or other software application on a client. In some embodiments, the provisioning and / or request of end user 202 can be provided to the ingress gateway 230. The ingress gateway 230 may include a corresponding agent to handle the provisioning and / or request for services (one or more services provided in the computing environment 201).

[0045] Services are provided or deployed based on various virtualization technologies supported by the computing environment 201. In some embodiments, services can be provided based on virtual machine (VM)-based virtualization, container-based virtualization, and / or similar methods. Virtual machine-based virtualization can simulate a real computer by initializing a virtual machine, executing programs and applications without directly contacting any actual hardware resources. While the virtual machine virtualizes the machine, according to container-based virtualization, a container can be started to virtualize the entire operating system (OS) so that multiple workloads can run on a single operating system instance.

[0046] In one embodiment based on container virtualization, several containers of a service can be assembled into a computing component (e.g., a Kubernetes Pod). For example, as shown in Figure 2, service 220-2 can be equipped with one or more computing components (Pods) Pod240-1, 240-2, ..., 240-N (collectively referred to as Pods). The Pod may include a proxy 245 and one or more containers 242-1, 242-2, ..., 242-M (collectively referred to as containers). One or more containers in the Pod process requests related to one or more corresponding functions of the service, and the proxy 245 generally controls network functions related to the service, such as routing, load balancing, etc. Other services can also be equipped with Pods similar to Pods.

[0047] During operation, executing a user request from end user 202 may require invoking one or more services in computing environment 201. Executing one or more functions of one service may require invoking one or more functions of another service. As shown in FIG2 , service "A" 220-1 receives a user request from end user 202 from ingress gateway 230. Service "A" 220-1 may invoke service "D" 220-2, and service "D" 220-2 may request service "E" 220-3 to execute one or more functions.

[0048] This computing environment can be a cloud computing environment, where resource allocation is managed by the cloud service provider, allowing for feature development without having to worry about implementing, adjusting, or scaling servers. This computing environment allows developers to execute code in response to events without building or maintaining complex infrastructure. Services can be partitioned to perform a set of functions that can scale independently and automatically, rather than scaling a single hardware device to handle the potential load.

[0049] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this disclosure, such as weather forecast results and other data, are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0050] In the above operating environment, the present disclosure provides a method for generating interactive information, which can be applied to a generative question-answering system, wherein the generative question-answering system can be used to answer user inquiries, such as chat robots, voice assistants, etc. This is only an example and does not specifically limit the type of generative question-answering system. Figure 3 is a flow chart of a method for generating interactive information according to an embodiment of the present disclosure. As shown in Figure 3, the method may include the following steps:

[0051] Step S302: Input inquiry information in the operation interface of the generative question answering system.

[0052] In the technical solution provided in step S302 above, a query can be input into the user interface of the generative question-answering system. The query can include questions, questions, or inquiries posed by a user on a terminal device. The generative question-answering system can be used to answer questions, can be a generative search link, or can be used to generate summaries or abstracts based on a large model. This is for illustrative purposes only and does not impose any specific limitations on the type of generative question-answering system.

[0053] Optionally, when a user wants to ask a question, he or she may input query information into the generative question answering system. The generative question answering system obtains the query information input by the user and displays it in the operation interface.

[0054] Step S304: Analyze the query information using the generative search model to generate initial response content that matches the query information.

[0055] In the technical solution provided in the above step S304 of the present disclosure, after obtaining the inquiry information, the inquiry information can be analyzed using a generative search model to obtain initial reply content that matches the inquiry information. Among them, the generative search model is a large model, which can be a large-scale pre-trained model based on deep learning, which can be used to determine the initial reply content, and can also be called agent one. The initial reply content can be a reply result or a reply framework for the inquiry information, and can include hallucination answers. The reply framework can be a reply idea for clearly answering the inquiry information, or it can be a logical structure or framework for the reply. This is only for illustration, and no specific restrictions are placed on the content of the reply framework.

[0056] Optionally, after obtaining the query information entered in the operation interface, the generative search model can be called up and used to analyze the query information. If the generative search model can determine the answer to the query information, the reply content of the query information can be directly generated. If the generative search model is not sure about the answer to the query information, a reasonable reply framework can be given.

[0057] Step S306: query and obtain retrieval information associated with the query information, wherein the retrieval information is used to at least represent the results retrieved based on the query information.

[0058] In the technical solution provided in step S306 of the present disclosure, the generative question-answering system can query for retrieval information associated with the query information. The retrieval information can be used to at least characterize the results retrieved based on the query information, can be used to guide the generative search model in generating response content, and can include noise information and background knowledge related to the query information. It can be information obtained from other locations such as articles, web pages, and news reports, and can be text information, audio information, etc. This is for illustrative purposes only and does not impose specific restrictions on the source and type of the query information.

[0059] Optionally, search information associated with the query information can be searched from papers, web pages, news, etc.

[0060] For example, assuming that the inquiry information is about diabetes, you can search for diabetes-related information from medical websites, health blogs, academic forums, etc., and use this search information to determine the accuracy of the initial reply content.

[0061] Step S308, obtaining associated reply content that is at least associated with the initial reply content and the search information, wherein the associated reply content is generated by performing at least one round of interactive reply using at least one auxiliary generative search model and a generative search model of the generative search model.

[0062] In the technical solution provided in the above step S308 of the present disclosure, the generative question-answering system includes two subsystems: a generative search model and at least one auxiliary generative search model. The generative question-answering system can respectively utilize at least one auxiliary generative search model and a generative search model, and the above subsystems perform at least one round of interactive replies to generate associated reply content that is at least associated with the initial reply content and the retrieved information. Among them, at least one auxiliary generative search model can also be a large-scale pre-trained model based on deep learning, which can also be called agent two. The associated reply content can be the content obtained by optimizing the initial reply content.

[0063] In this embodiment, multiple auxiliary generative search models can be set up according to actual needs, and at least one round of interactive replies can be performed using the auxiliary generative search models and the generative search model to obtain associated reply content. The associated reply content is used to process the initial reply content generated by the generative search information to improve the accuracy of the reply content finally generated by the generative question-answering system.

[0064] Optionally, at least one auxiliary generative search model obtains the initial reply content generated by the generative search model and the retrieved retrieval information, and performs at least one round of interactive reply with the generative search model based on the initial reply content and the retrieval information to generate associated reply content.

[0065] For example, consider the query "How can I prevent diabetes?" entered into the user interface. The generative search model is then invoked to analyze the query and generate an initial response: "We recommend that you take the following measures to prevent diabetes: maintain a healthy diet and exercise regularly." Diabetes-related search information is retrieved from medical websites, health blogs, academic forums, and other locations. At least one auxiliary generative search model processes the search information and initial response information and transmits the processed results to the generative search model. This process then uses the auxiliary and generative search models to perform at least one round of interactive responses, ultimately yielding relevant responses.

[0066] Step S310: Replace the initial reply content with the associated reply content to obtain the target reply content that matches the query information.

[0067] In the technical solution provided in step S310 above, after obtaining the associated reply content, the initial reply content can be replaced with the associated reply content to obtain the target reply content that matches the query information. The target reply content can be the final reply result, which can be the final output of the generative question-answering system.

[0068] In this embodiment, it is taken into account that when the generative search model determines the reply content of the query information, if it is disturbed by the retrieval information, it will obtain an erroneous reply. In order to avoid the above problem and to improve the accuracy of the target reply content generated by the generative question-answering system, in this embodiment, at least one auxiliary generative search model is set, and at least one auxiliary generative search model collaborates with the generative search model. After obtaining the query information, the retrieval information is ignored, and the generative search model only analyzes the query information to obtain the initial reply content that matches the query information. At this time, the initial reply content is the content obtained without considering the retrieval information. Query the retrieval information associated with the query information. At least one auxiliary generative search model processes the retrieval information and the initial reply content, and the processed result can be adjusted by the generative search model based on the initial reply content. In this way, the auxiliary generative search model and the generative search model perform at least one round of interactive reply to obtain associated reply content associated with the initial reply content and the retrieval information. The initial reply content is replaced with the associated reply content to obtain the target reply content that matches the query information, thereby achieving the purpose of improving the accuracy of the target reply content and solving the technical problem of low generation accuracy of target interactive information.

[0069] As can be seen from the above, the method utilizes at least one auxiliary generative search model and a generative search model to perform at least one round of interactive responses. When the initial response content is a hallucinatory answer, at least one auxiliary generative search model judges the initial response content based on the retrieval information, and can determine that the initial response content conflicts with the real information in the retrieval information. On this basis, at least one auxiliary generative search model can adjust the initial response content to obtain associated response content. Based on the associated response content, the target response content that matches the query information can be obtained, thereby avoiding the problem of the target response content conflicting with the real information. Since the target response content is obtained through multiple rounds of collaboration and editing, the technical problem of low response accuracy due to misleading retrieval information or answer hallucinations is solved, and the technical effect of improving the accuracy of the response is achieved.

[0070] Optionally, after obtaining the associated reply content, it may be determined whether the associated reply content and the initial reply content express the same meaning, whether there is any ambiguity, or whether the content in the associated reply content is correct or reasonable. If the associated reply content is correct, the initial reply content may be replaced with the associated reply content to obtain the target reply content that matches the query information.

[0071] For example, suppose a user enters the query "How to prevent a cold?" into the generative question-answering system's interface. The query is then retrieved. The generative search model analyzes the query and generates an initial response: "Methods for preventing a cold include drinking plenty of water, washing hands frequently, and avoiding going out for fresh air." The generative question-answering system then retrieves search information related to the query, such as advice and methods for preventing colds from medical websites. Based on the search information and the initial response, it generates a related response: "Methods for preventing a cold include going out for fresh air and drinking plenty of water." This indicates that the initial response may contain errors. In this case, the initial response can be replaced with the related response, resulting in a target response that matches the query: "Methods for preventing a cold include going out for fresh air and drinking plenty of water."

[0072] In this embodiment, a generative search model based on an agent community is proposed, in which multiple types of agents are set up in the community, for example, a generative search model and at least one auxiliary generative search model. The retrieval information and the initial reply content are edited multiple times by multiple types of agents to achieve the technical effect of improving the accuracy of the answer to the question and solve the technical problem of low accuracy of the answer to the question.

[0073] In the disclosed embodiment, an auxiliary generative search model is proposed, which uses the generative search model to first analyze the query information to obtain the initial reply content, and then queries to obtain the retrieval information associated with the query information. The auxiliary generative search model and the generative search model interact to generate the associated reply content, and the initial reply content is updated to obtain the final target reply content. The initial reply content generated by the generative search model is adjusted through the associated reply content, thereby achieving the technical effect of improving the accuracy of the generated interactive information and solving the technical problem of low accuracy of the generated interactive information.

[0074] The above method of this embodiment is further introduced below.

[0075] As an optional implementation, step S304 uses a generative search model to analyze the query information and generate initial response content that matches the query information, including: determining first prompt information associated with the query information, wherein the first prompt information is used to represent guidance information for obtaining the initial response content; and guiding the generative search model to generate the initial response content according to the first prompt information.

[0076] In this embodiment, first prompt information associated with the query information can be determined, the first prompt information can be input into a generative search model, and the generative search model can be guided by the first prompt information to generate an initial response content for the query information. That is, the initial response content is the output data of the generative search model. The first prompt information can be an initial prompt information (prompt), which can be used to represent guidance information for obtaining the initial response content. For example, it can be an operating instruction (Instruction) in a large model, or it can be a keyword, phrase, or context information related to the query information.

[0077] Optionally, the first prompt information can also be used in the generative search model to guide the user to start interacting with the generative search model, and can also provide some basic information and examples to help the user better understand how to communicate with the generative search model and ask questions. The first prompt information can also include the capabilities and limitations of the generative search model, how to ask questions to obtain results that meet the requirements, how to deal with vague or incomplete questions, etc. Through these first prompt information, users can communicate more efficiently using the generative search model to obtain the information they need. It should be noted that this is only an example, and there is no specific restriction on the type and content of the first prompt information.

[0078] For another example, when a query (Q) is received, the first prompt information associated with the query information can be determined: "Please help me formulate an accurate, engaging, and concise Chinese answer to the given query information. You should strive to answer this question using your internal knowledge. If not, you should provide a response framework that is unbiased and uses a journalistic tone" and "Note that the answer needs to be in Chinese." Based on the first prompt information, the generative search model analyzes the query information Q to generate the initial response content (Last_A) of the Chinese answer.

[0079] As an optional implementation, according to the first prompt information, guiding the generative search model to generate initial reply content, including: according to the first prompt information, guiding the generative search model to determine the initial answer as the initial reply content when the initial answer to the query information is successfully generated; according to the first prompt information, guiding the generative search model to generate a reply framework that matches the query information when the initial answer to the query information fails to be generated, and determining the reply framework as the initial reply content, wherein the reply framework at least includes the logic for generating the initial answer.

[0080] In this embodiment, the generative search model can be guided to generate initial response content according to the first prompt information. If the generative search model successfully generates an initial answer to the query information, the initial answer can be determined as the initial response content. If the generation of the initial response fails, a response framework matching the query information can be generated and determined as the initial response content. The response framework can at least include the logic for generating the initial answer, and can be an answer framework or a reply framework, which can be used to provide a response idea for the query information.

[0081] Optionally, after receiving the query, the generative search model analyzes the query according to the first prompt and attempts to generate an initial answer. If the initial answer is successfully generated, the initial answer is directly determined as the initial reply content. If the initial answer fails to be generated, a reply framework matching the query is generated, where the reply framework at least includes the logic for generating the initial answer, and the reply framework is then determined as the initial reply content.

[0082] For example, consider a query like "What's the weather like today?" If the generative search model successfully generates an initial answer: "It's sunny today," the initial answer is used as the initial response. If the generative search model fails to generate an initial response, it can generate a response framework: "The weather today is [description]," which is then used as the initial response.

[0083] As an optional implementation, step S308, obtaining associated reply content that is at least associated with the initial reply content and the retrieval information, includes: using an auxiliary generative search model to analyze the retrieval information and the initial reply content to obtain an analysis result, wherein the analysis result is used to characterize the degree of matching between the initial reply content and the retrieval information; based on the analysis result, controlling the auxiliary generative search model to generate associated reply content.

[0084] In this embodiment, the assisted generative search model can be used to analyze the search information and the initial response content to determine whether the search information and the initial response content match, thereby obtaining an analysis result. Based on the analysis result, the assisted generative search model can be further controlled to generate related response content. The analysis result can be used to indicate the degree of match between the initial response content and the search information.

[0085] Optionally, the auxiliary generative search model obtains the initial response content of the generative search model and simultaneously sees the query information and the search information related to the query information. The auxiliary generative search model examines the output of the generative search model and determines the degree of match between the initial response content output by the generative search model and the search information to obtain an analysis result. Based on the analysis result, the auxiliary generative search model can be controlled to generate related response content.

[0086] Optionally, if the analysis results show a high degree of match, it indicates that the initial response content is highly relevant to the search information; if the degree of match is low, it indicates that the initial response content is less relevant to the search information. For example, if the user enters the query "How to cook pasta," and the initial response content is "The steps for cooking pasta are to boil the noodles first." The query results in the search for the cooking steps for pasta, and the auxiliary generative search model can be used to analyze the degree of match between the search information and the initial response content. After analysis, the result is a high degree of match, indicating that the initial response content is relevant to the search information. In this case, the auxiliary generative search model can generate more relevant related response content based on the analysis results. The related response content may be "The steps for cooking pasta are to boil the noodles first and then serve with pasta sauce."

[0087] As an optional implementation, based on the analysis results, the auxiliary generative search model is controlled to generate associated reply content, including: if the analysis result is that the initial reply content fails to match the retrieval information, the auxiliary generative search model is controlled to adjust the initial reply content; in at least one auxiliary generative search model, when the auxiliary generative search model has a next auxiliary generative search model, the next auxiliary generative search model is used to analyze the adjusted initial reply content to obtain associated reply content.

[0088] In this embodiment, if the analysis result indicates that the initial response content fails to match the search information, the auxiliary generative search model can be controlled to adjust the initial response. If a next auxiliary generative search model exists in at least one auxiliary generative search model, the adjusted initial response content is adjusted using the next auxiliary generative search model to obtain associated response content. The next auxiliary generative search model can be a generative search model, another auxiliary generative search model, or another proxy.

[0089] Optionally, if the analysis result indicates that the initial reply content fails to match the search information, it can be determined that the initial reply content is inconsistent with the search information, and the auxiliary generative search model can be controlled to adjust the initial reply content. If a next auxiliary generative search model exists in the auxiliary generative search model, the next auxiliary generative search model can be used to analyze the adjusted initial reply content to obtain associated reply content.

[0090] In this embodiment, the number of auxiliary generative search models can be set according to actual needs. For example, if you want to obtain target response content with a higher accuracy, you can set a larger number of auxiliary generative search models to perform multiple rounds of interactive replies. If you want to obtain the target response content in a shorter time, you can set a smaller number of auxiliary generative search models. After the generative search model obtains the query information, it can generate initial response content that matches the query information. Retrieve the retrieval information associated with the query information, and transmit the initial response content and the retrieval information to the auxiliary generative search model. The auxiliary generative search model analyzes the degree of match between the retrieval information and the initial response content to obtain an analysis result. If the analysis result shows that the initial response content fails to match the retrieval information, the auxiliary generative search model can be controlled to adjust the initial response content. In the case of multiple auxiliary generative search models, the adjusted initial response content can be transmitted to the next auxiliary generative search model, and the next auxiliary generative search model is used to further adjust the initial response content to obtain the final associated response content.

[0091] For example, the generative search model can be used as the second auxiliary generative search model. After the first auxiliary generative search model adjusts the initial response content output by the generative search model, the adjusted initial response content is transmitted to the second auxiliary generative search model, and the second auxiliary generative search model is used to further adjust the adjusted initial response content through prompt information to obtain the final associated response content. Alternatively, multiple different auxiliary generative search models can be set up. After the first auxiliary generative search model obtains the initial response content output by the generative search model, it can adjust the initial response content based on the retrieval information and output the adjusted initial response content to the second auxiliary generative search model. The second auxiliary generative search model can further adjust the adjusted initial response content based on the retrieval information to obtain the associated response content.

[0092] As an optional implementation, in at least one auxiliary generative search model, when there is a next auxiliary generative search model in the auxiliary generative search model, the next auxiliary generative search model is used to analyze the adjusted initial reply content to obtain the associated reply content, including: a first determination step, when there is a next auxiliary generative search model in the auxiliary generative search model in at least one auxiliary generative search model, the next auxiliary generative search model is used to determine whether the adjusted initial reply content successfully matches the query information; a second determination step, if the next auxiliary generative search model is used to determine that the adjusted initial reply content successfully matches the query information, the next auxiliary generative search model is controlled to determine the adjusted initial reply content as the associated reply content; a third determination step, if the next auxiliary generative search model is used to determine that the adjusted initial reply content fails to match the query information, the next auxiliary generative search model is controlled to adjust the adjusted initial reply content to obtain an adjustment result, and the next auxiliary generative search model is determined as the auxiliary generative search model, the adjustment result is determined as the adjusted initial reply content, and the first determination step is returned to be executed until the adjusted initial reply content successfully matches the query information, or the number of times the first determination step is executed reaches a threshold number.

[0093] In this embodiment, when the auxiliary generative search model has a next auxiliary generative search model in at least one auxiliary generative search model, the adjusted initial reply result can be output to the next auxiliary generative search model. The next auxiliary generative search model determines whether the adjusted initial reply content matches the query information successfully. If the next auxiliary generative search model determines that the adjusted initial content matches the query information successfully, the adjusted initial reply content can be determined as the associated reply content. If the adjusted initial reply content fails to match the query information, the next auxiliary generative search model can be controlled to adjust the adjusted initial reply content to obtain an adjustment result, and the next auxiliary generative search model is determined as the auxiliary generative search model, and the adjustment result is determined as the adjusted initial reply content, and the initial reply content is further adjusted until the final adjusted initial reply content matches the query information successfully, and the adjusted initial reply content is determined as the associated reply content, or when the number of adjustments reaches a threshold, the adjustment can be stopped, and the final adjusted initial reply content before the adjustment is stopped is determined as the associated reply content.

[0094] Optionally, the generative search model processes the query to generate an initial response. The generative question-answering system determines the search information associated with the query and transmits the search information and the initial response to the auxiliary generative search model. The auxiliary generative search model determines the degree of match between the search information and the initial response to generate an analysis result.

[0095] If the analysis result is used to characterize the failure of the match between the retrieval information and the initial reply content, the auxiliary generative model adjusts the initial reply content and transmits the adjusted initial reply content to the next auxiliary generative search model, where the next auxiliary generative search model can be a generative search model. The next auxiliary generative search model determines the matching relationship between the adjusted initial reply content and the query information. If the match is successful, the next auxiliary generative search model can output "Exit Editing" to indicate that the reply is completed, and the adjusted initial reply content is determined as the associated reply content. If the match fails, the current auxiliary generative search model can further optimize and adjust the adjusted initial reply content, and transmit the further optimized and adjusted initial reply content to the next auxiliary generative search model of the next auxiliary generative search model, and use the auxiliary generative search model to determine the degree of matching between the retrieval information and the adjusted initial reply content, and execute it cyclically according to the above adjustment method until a certain auxiliary generative search model outputs "Exit Editing", then the adjusted initial reply content in the auxiliary generative search model can be determined as the associated reply content, or, if the number of executions in the above method exceeds the pre-set number of iterations, the last adjusted initial reply content will be determined as the associated reply content.

[0096] For example, assume that the generative search model is Agent 1, and the auxiliary generative search model is Agent 2, and Agent 1 can also serve as the auxiliary generative search model. After receiving the query information, Agent 1 analyzes the query information and determines the initial response content. The generative question-answering system queries the search information associated with the query information and transmits the search information and the initial response content to Agent 2. Agent 2 determines the consistency between the search information and the initial response content. If they are inconsistent, Agent 2 adjusts the initial response content and transmits the adjusted initial response content to Agent 1. Agent 1 receives Agent 2's edited initial response content and analyzes the adjusted initial response content based on the query information. If the query information matches the adjusted initial response content, Agent 1 can output the adjusted initial response content as the associated response content.

[0097] If the inquiry information does not match the adjusted initial reply content, Agent 1 can further adjust the adjusted initial reply content based on the inquiry information, and transmit the adjusted initial reply content to Agent 2. Agent 2 can further judge the secondary adjusted initial reply content based on the retrieval information. If they match, the secondary adjusted initial reply content can be used as the associated reply content. If they do not match, the secondary adjusted initial reply content can be adjusted again. The above steps are iterated until Agent 1 outputs "Exit Editing", and the last adjusted initial reply content is determined as the associated reply content, or until the iteration reaches the predetermined number of iterations, and the final adjusted initial reply content can be determined as the associated reply content.

[0098] It should be noted that, in addition to agent 2, the auxiliary generative search model can also be other agents besides agents 1 and 2, such as agent 3, etc. This is only an example and does not impose specific restrictions on the type of auxiliary generative search model.

[0099] This embodiment proposes a collaborative agent approach. Through the coordinated operation of Agent 1 and Agent 2, the accuracy and rationality of responses are gradually improved, thereby achieving the technical effect of improving the accuracy of the target responses generated by the generative question-answering system and resolving the technical issue of low accuracy in generating target responses. It should be noted that the number and type of Agent 1 and Agent 2 can be selected based on actual needs.

[0100] As an optional implementation, if the analysis result is that the initial reply content fails to match the retrieval information, the auxiliary generative search model is controlled to adjust the initial reply content, including: if the analysis result is that the initial reply content fails to match the retrieval information, second prompt information associated with the retrieval information is determined, wherein the second prompt information is used to represent the guiding information for adjusting the initial reply content; according to the second prompt information, the auxiliary generative search model is guided to adjust the initial reply content.

[0101] In this embodiment, an auxiliary generative search model is used to analyze the degree of match between the initial reply content and the retrieval information to obtain an analysis result. If the analysis result is that the initial reply content fails to match the retrieval information, second prompt information associated with the retrieval information can be determined, and according to the second prompt information, the auxiliary generative search model is guided to adjust the initial reply content. Among them, the second prompt information can be an optimization prompt, which can be used to represent guiding information for adjusting the content of the initial answer, or it can be pre-set information, for example, it can be "You are a helpful question answering assistant, and you are working with me to answer the given question. We need to develop accurate, attractive and concise Chinese answers for the given question. After searching online, I have combined the search results and modified your previous answer into the candidate below. However, my answer may be too complicated and contain some irrelevant content. Your goal is to make a concise and relevant answer to the question with quotations, so you can only delete irrelevant sentences in the answer to the given question. Please note that you should not change the meaning of my answer, especially for sentences with citation or deleted quote marks. But for sentences with more than 3 quotations, you must delete them. Note that you should answer in Chinese!" It should be noted that this is only an example and does not specifically limit the type and determination method of the second prompt information.

[0102] Optionally, if the auxiliary generative search model analyzes the initial response content and the search information and determines that the initial response content fails to match the search information, the generative question-answering system can determine second prompt information associated with the search information. The second prompt information can be used to guide the next auxiliary generative search model of the auxiliary generative search model to adjust the adjusted initial response content.

[0103] As an optional implementation, according to the second prompt information, guiding the auxiliary generative search model to adjust the initial reply content, including: according to the second prompt information, guiding the auxiliary generative search model to determine the target content allowed to be modified in the initial reply content, and adjusting the target content allowed to be modified in the initial reply content.

[0104] In this embodiment, during the process of guiding the auxiliary generative search model to adjust the initial response content according to the second prompt information, the auxiliary generative search model can be guided based on the second prompt information to determine target content in the initial response content that is permitted to be modified, and the replicated generative search model can be controlled to adjust the target content in the initial response content that is permitted to be modified. The target content can be sentences, words, or other content in the adjusted initial response content that is irrelevant to the inquiry information. This is for illustrative purposes only and does not specifically limit the type of target content.

[0105] Optionally, the second prompt information can limit the target content that is allowed to be modified in the initial reply content, such as phrases and sentences that are allowed to be deleted in the initial reply content. After obtaining the initial reply content, the auxiliary generative search model can be controlled to modify the target content in the initial reply content.

[0106] As an optional implementation, the next auxiliary generative search model includes a generative search model.

[0107] In this embodiment, the next auxiliary generative search model may include a generative search model and other auxiliary generative search models, and no specific limitation is imposed on the type of the auxiliary generative search model in the at least one auxiliary generative search model.

[0108] Optionally, Agent 1 can receive the query information and Agent 2's adjusted initial response, while Agent 2 can receive the search information, the initial response, and the query information. Agent 1 can generate the initial response based solely on the query information, thereby avoiding interference from the search information. Agent 2 can view the search information and Agent 1's initial response and adjust Agent 1's initial response to avoid conflicts between the initial response and the search information, thereby improving the accuracy of the final target response.

[0109] As an optional implementation, the method may further include: if the analysis result shows that the initial reply content successfully matches the search information, determining the initial reply content as the target reply content.

[0110] In this embodiment, the auxiliary generative search model can match the retrieval information and the initial reply content to obtain an analysis result. If the analysis result is that the initial reply content successfully matches the retrieval information, it means that the initial reply content is not an hallucination answer and does not conflict with the factual information. Therefore, the initial reply content can be determined as the target reply content.

[0111] As an optional implementation, querying to obtain retrieval information associated with the query information includes: querying a database using an auxiliary generative search model to obtain retrieval information associated with the query information.

[0112] In this embodiment, the generative question-answering system can also use the auxiliary generative search model to query the database to obtain retrieval information associated with the query information. The database can include data storage centers in multiple platforms, such as databases containing different types of news, papers, and videos. This is for illustrative purposes only and does not impose any specific restrictions on the type of database.

[0113] For example, when someone enters the query "I want to know more about giraffes," the assisted generative search model can be used to convert this query into a database query. According to the query, detailed information related to giraffes, such as their habits, hobbies, and species, can be retrieved from the database.

[0114] In the embodiment of the present disclosure, an auxiliary generative search model is proposed, which uses the generative search model to first analyze the query information. After obtaining the initial reply content, the query information is queried to obtain the retrieval information associated with the query information, and the associated reply content generated by the interactive reply between the auxiliary generative search model and the generative search model is used to replace the initial reply content to obtain the final target reply content, thereby achieving the technical effect of improving the accuracy of the generated interactive information and solving the technical problem of low accuracy of the generated interactive information.

[0115] The disclosed embodiments also provide another method for generating interactive information from the human-computer interaction side, which can be applied to a generative question-answering system deployed in a search engine. FIG4 is a flowchart of another method for generating interactive information according to an embodiment of the disclosed embodiments. As shown in FIG4 , the method may include the following steps.

[0116] Step S402: Display the query information on the operation interface of the generative question answering system.

[0117] In the technical solution provided in step S402 of the present disclosure, query information can be input into the operation interface of the generative question-answering system, wherein the query information can be the user's question information, question, query content, etc.

[0118] Optionally, when a user wants to ask a question, he or she may input query information into the search engine. The generative question answering system in the search engine may obtain the query information input by the user and input the query information into the operation interface.

[0119] Step S404, in response to the initial search instruction on the operation interface, initial response content matching the query information is displayed on the operation interface, wherein the initial response content is obtained by analyzing the query information using a generative search model.

[0120] In the technical solution provided in step S404 of the present disclosure, after a user enters a query, an initial search instruction can be issued by triggering a control on the operation interface. In response to the initial search instruction on the operation interface, an initial response matching the query can be displayed on the operation interface. The initial response can be obtained by analyzing the query using a generative search model.

[0121] Step S406, in response to the re-search instruction on the operation interface, the target reply content matching the query information is displayed on the operation interface, wherein the target reply content is obtained by replacing the initial reply content with the associated reply content, the retrieval information is used to at least characterize the results retrieved based on the query information, and the associated reply content is generated by performing at least one round of interactive reply using at least one auxiliary generative search model of the generative search model and the generative search model, and is associated with the initial reply content and the retrieval information.

[0122] As an optional embodiment, after determining the initial response content, the initial response content can be displayed on the operation interface. If the user feels that the accuracy of the initial interactive information generated is low, a re-search instruction can be initiated. In response to the re-search instruction acting on the operation interface, the search engine can control at least one auxiliary generative search model and the generative search model to perform at least one round of interactive reply to generate associated response content, and use the associated response content associated with the initial response content and the search information to update the initial response content to obtain the target response content. The search engine can display the target response content that matches the query information on the operation interface.

[0123] As an optional implementation, the generative search model is a pre-trained model based on deep learning.

[0124] In this embodiment, the generative search model may be a pre-trained model obtained through local deep learning, which may also be referred to as a large model.

[0125] As an optional embodiment, the inquiry information is multimodal information, and the type of the multimodal information includes at least one of the following: text information containing character information, video frame information, and audio information; the type of the target reply content includes at least one of the following: text information, image information, video information, and voice information.

[0126] In this embodiment, the query information can be multimodal information, including at least text information including character information, video frame information, audio information, etc. The types of target response content can include at least text information, image information, video information, and voice information. It should be noted that this is for illustrative purposes only and does not impose any specific limitations on the form of the query information and target response content.

[0127] In an embodiment of the present disclosure, query information is displayed on the operation interface of the generative question-answering system; in response to an initial search instruction on the operation interface, initial reply content matching the query information is displayed on the operation interface, wherein the initial reply content is obtained by analyzing the query information using a generative search model; in response to a second search instruction on the operation interface, target reply content matching the query information is displayed on the operation interface, wherein the target reply content is obtained by replacing the initial reply content with associated reply content, and the retrieval information is used to at least characterize the results retrieved based on the query information, and the associated reply content is generated by performing at least one round of interactive replies using at least one auxiliary generative search model and a generative search model of the generative search model, and is associated with the initial reply content and the retrieval information, thereby achieving the technical effect of improving the accuracy of the generated interactive information and solving the technical problem of low accuracy of the generated interactive information.

[0128] According to an embodiment of the present disclosure, another method for generating interactive information is provided in the context of Software as a Service (SAAS). This method can be applied to a generative question-answering system. FIG5 is a flow chart of another method for generating interactive information according to an embodiment of the present disclosure. As shown in FIG5, the method may include the following steps:

[0129] Step S502: Input inquiry information by calling a first interface, wherein the first interface includes a first parameter, and the parameter value of the first parameter is the inquiry information.

[0130] In the technical solution provided in step S502 of the present disclosure, the input query information can be obtained by calling a first interface, wherein the first interface may include a first parameter, and the parameter value of the first parameter may be the query information. The query information may be information such as a question raised by the user, and may be multimodal information.

[0131] For example, a user can pass query information as the parameter value of the first parameter through the Application Programming Interface (API) interface of the SAAS service provider to query or process the query information. The SAAS service provider's system will receive the query information and process the query information. The first interface can be an API endpoint provided by the SAAS platform. The user can send the query information by calling this interface, and the first parameter is used to pass the specific query information. In this way, the user can use the functions provided by the SAAS platform to perform customized information query and processing.

[0132] Step S504: Analyze the query information using the generative search model to generate initial response content that matches the query information.

[0133] Step S506: query and obtain retrieval information associated with the query information, wherein the retrieval information is used to at least represent the results retrieved based on the query information.

[0134] Step S508, obtaining associated reply content that is at least associated with the initial reply content and the search information, wherein the associated reply content is generated by performing at least one round of interactive reply using at least one auxiliary generative search model and a generative search model of the generative search model.

[0135] Step S510: Replace the initial reply content with the associated reply content to obtain target reply content that matches the query information.

[0136] Step S512: Output the target reply content by calling the second interface, wherein the second interface includes a second parameter, and the parameter value of the second parameter is the target reply content.

[0137] In the technical solution provided in the above step S512 of the present disclosure, the target reply content can be output by calling the second interface, wherein the second interface can include a second parameter, and the parameter value of the second parameter can be the target reply content.

[0138] In an embodiment of the present disclosure, query information is input by calling a first interface, wherein the first interface includes a first parameter, and the parameter value of the first parameter is the query information; the query information is analyzed by using a generative search model to generate initial reply content that matches the query information; retrieval information associated with the query information is obtained by querying, wherein the retrieval information is used to at least represent the results retrieved based on the query information; associated reply content associated with at least the initial reply content and the retrieval information is obtained, wherein the associated reply content is generated by executing at least one round of interactive reply using at least one auxiliary generative search model and a generative search model of the generative search model; the initial reply content is replaced with the associated reply content to obtain target reply content that matches the query information; the target reply content is output by calling a second interface, wherein the second interface includes a second parameter, and the parameter value of the second parameter is the target reply content, thereby achieving the technical effect of improving the accuracy of the generated interactive information and solving the technical problem of low accuracy of the generated interactive information.

[0139] According to an embodiment of the present disclosure, an embodiment of a generative question-answering system is also provided. FIG6 is a schematic diagram of a generative question-answering system according to an embodiment of the present disclosure. As shown in FIG6 , the generative question-answering system 600 may include: a first information generation terminal 602 and a second information generation terminal 604.

[0140] The first information generating terminal 602 is configured to obtain query information input into the operation interface of the generative question-answering system; analyze the query information using the generative search model, and generate initial response content matching the query information.

[0141] In this embodiment, the first message generating terminal 602 may include a generative search model, which may be used to input query information in the operation interface of the generative question-answering system, and analyze the query information using the generative search model to obtain initial response content that matches the query information.

[0142] Optionally, the first message generating end 602 may include an interactive end of a generative search model. After obtaining the query information, the generative search model may be used to analyze the query information to obtain the initial reply content.

[0143] The second information generation terminal 604 is configured to perform at least one round of interactive response using at least one auxiliary generative search model of the generative search model and the generative search model to generate associated response content that is at least associated with the initial response content and the retrieval information of the query information, wherein the retrieval information is used to at least represent the results retrieved based on the query information.

[0144] In this embodiment, the second information generating terminal 604 may include at least one auxiliary generative search model based on the generative search model, and may obtain the search information and the initial response content output by the first information generating terminal 602. After obtaining the search information and the initial response content, at least one round of interactive response may be performed using the at least one auxiliary generative search model and the generative search model to generate associated response content associated with the initial response content and the search information of the query information.

[0145] In this embodiment, the first information generating terminal 602 may be configured to replace the initial reply content with the associated reply content, and output the target reply content that matches the inquiry information.

[0146] In this embodiment, the first information generating end 602 and the second information generating end 603 may be deployed on the same end or on different ends, and no specific limitation is made here.

[0147] For example, the first information generating terminal 602 obtains query information and analyzes the query information using a generative search model to generate initial response content that matches the query information. The initial response content is then transmitted to the second information generating terminal 602. The second information generating terminal 602 obtains search information and the initial response content, and adjusts the initial response content based on the search information to obtain adjusted initial response content. The adjusted initial response content can then be transmitted to the first information generating terminal 602. The first information generating terminal 602 obtains the adjusted initial response content and further determines the adjusted initial response content based on the query information. If the query information and the adjusted initial response content are consistent, the adjusted initial response content is determined as the associated response content and replaced with the associated response content to output the target response content that matches the query information. If they are inconsistent, the adjusted initial response content can be further adjusted and transmitted to the second information generating terminal. Following the aforementioned steps, the initial response content is iteratively updated again to obtain the final target response content.

[0148] In this embodiment, query information input into the operation interface of the generative question-answering system is obtained through the first information generation terminal 602; the query information is analyzed using the generative search model to generate initial reply content matching the query information; and at least one round of interactive reply is performed through the second information generation terminal 604 using at least one auxiliary generative search model of the generative search model and the generative search model to generate associated reply content associated with at least the initial reply content and the retrieval information of the query information, wherein the first information generation terminal is used to replace the initial reply content with the associated reply content, and output the target reply content matching the query information, thereby achieving the technical effect of improving the accuracy of the generated interactive information and solving the technical problem of low accuracy of the generated interactive information.

[0149] Currently, generative search is used in scenarios where a large model combines search information with responses. In these scenarios, large models can sometimes generate inaccurate responses to the target interaction information. For example, for some questions, the large model possesses relevant knowledge and can answer correctly, but can be misled by the retrieved information and answer incorrectly. Alternatively, the large model lacks relevant knowledge and, without reference to the provided search information, may generate incorrect responses that are hallucinated.

[0150] To solve the above problems, in this embodiment, a generative search model based on an agent community is proposed. Multiple types of agents are set up in the community, such as a generative search model and at least one auxiliary generative search model. The retrieval information and the initial reply content are edited multiple times by multiple types of agents to achieve the technical effect of improving the accuracy of the answer to the question and solve the technical problem of low accuracy of the answer to the question.

[0151] The following takes two agents as an example to further introduce a large-model generative search method based on an agent community proposed in an embodiment of the present disclosure.

[0152] In this embodiment, it is assumed that the solution based on the agent community mainly has two agents, among which Agent 1 can only see the query information raised by the user and the answer generated by Agent 2, that is, the adjusted initial reply content; Agent 2 can see the query information, retrieval information and the initial reply content generated by Agent 1.

[0153] In this embodiment, since Agent 1 does not see the search information, it can avoid being misled by the search information when generating the initial response. It can also provide a reasonable response framework for questions that the large model does not know about. Furthermore, since Agent 2 can see the search information and Agent 1's response, if Agent 1 generates a hallucinated answer, Agent 2 can determine whether the initial response generated by Agent 1 conflicts with the retrieved factual information. Based on this, Agent 2 can adjust the initial response to achieve the accurate target response, alleviating the problem of hallucinated answers. Furthermore, Agent 2 can organize the search information more systematically based on the response framework provided by Agent 1, making it easier to analyze and arrive at the correct answer.

[0154] FIG7 is a flow chart of a large model generative search method based on an agent community according to an embodiment of the present disclosure. As shown in FIG7 , the method may include the following steps:

[0155] Step S701: The session starts and the user's query information is obtained.

[0156] In this embodiment, a conversation begins, the generative answer system receives the user's query information, and retrieves relevant retrieval information.

[0157] Step S702: Agent 1 processes the inquiry information and obtains the initial response content.

[0158] In this embodiment, agent one is responsible for answering the user's inquiry information based on the first prompt information. If agent one can answer the inquiry information correctly, the initial reply content is directly generated; otherwise, a reply framework can be generated to provide a basic structure or idea.

[0159] For example, suppose the first prompt information is: "Please help me formulate an accurate, attractive, and concise Chinese answer to the given inquiry information. You should try to answer this question with your internal knowledge. If not, you should give me a response framework with an unbiased and journalistic tone" and "Note that the answer needs to be in Chinese", the first prompt information and the inquiry information are input into the generative search model to obtain the initial response content.

[0160] Step S703: Agent 2 analyzes the initial reply content.

[0161] In this embodiment, Agent 2 receives Agent 1's initial response, which may include a reply or a reply framework, as well as query information and search information related to the query. Agent 2 then checks Agent 1's initial response to determine whether it matches the search information. If Agent 1's initial response is inconsistent with the search information, Agent 2 can modify it to generate a more accurate response. If they are consistent, the initial response can be provided to the user as the target response, for example, by displaying it on a user interface.

[0162] Optionally, after the editing is completed, Agent 2 can pass the adjusted initial response content to Agent 1 for further processing.

[0163] Optionally, the second agent may analyze the search information and the initial reply content based on the second prompt information, wherein the second prompt information may be an editing prompt.

[0164] For example, assuming the second prompt information is:

[0165] Please at least help me formulate an accurate and engaging Chinese response to the given inquiry information. The given inquiry information and my initial response content are as follows. However, for some examples, my initial response content may be useless, biased, untrue, and overly detailed. However, the following search information is provided (some of which may be irrelevant to the inquiry information).

[0166] Request information: {Q};

[0167] Initial reply content: {Last_A};

[0168] Retrieve information: The following doc from the search results is used to reference {d};

[0169] After reviewing the query information, the initial response content, and the search information, please provide your updated response (i.e., the adjusted initial response content). Note that your updated response should include citations from the search information below and fully address the question. Use a non-biased and journalistic tone. Always cite any factual claims and append citations after each sentence. When citing multiple search information, use labels, such as [1][2][3] to distinguish them. Cite at least one document and a maximum of three documents per sentence. If multiple documents support the sentence, cite only a subset of the documents. Note that you should answer in Chinese!

[0170] Action Token: "Exit Edit" ends the discussion and reaches consensus if the final answer is not used".

[0171] It should be noted that this is only an example, and the contents of the first prompt information and the second prompt information may be changed according to changes in actual conditions or changes in usage links.

[0172] Step S704: Agent 1 obtains the adjusted initial reply content from Agent 2 and further optimizes it.

[0173] In this embodiment, Agent 1 receives Agent 2's edited initial reply and compares it with the optimization prompt, namely, the second prompt information. Agent 1 then determines whether Agent 2's modifications to its initial reply are reasonable. If so, Agent 1 outputs "Exit Editing" to indicate the reply is complete and outputs the final target reply. If Agent 1 deems further optimization necessary, it makes modifications and retransmits the revised reply to Agent 2.

[0174] For example, in the current session, the second prompt is: You are a helpful question answering assistant, and you are working with me to answer the given inquiry information. We need to develop an accurate, attractive and concise Chinese answer to the given inquiry information. After searching online, I have combined the initial answer content based on the inquiry information with the searched retrieval information, and modified your previous initial answer content into the following candidate. However, my answer (that is, the adjusted initial answer content) may be too complicated and contain some irrelevant content. Your goal is to make a concise and relevant answer to the question with citations, so you can only delete irrelevant sentences in the answer to the given question. Please note that you should not change the meaning of my answer, especially for sentences with citation or dereference marks (for example, [1], [2], [3], etc.). But for sentences with more than 3 citations, you must delete them.

[0175] Request information: {Q};

[0176] Adjusted initial answer content: {My_answer_candidates};

[0177] Your updated answer:

[0178] Note that you should answer in Chinese;

[0179] Action Token: Exit editing, end discussion and reach consensus if final answer is not used.

[0180] Step S705: perform an iterative loop based on the above arrangement to obtain the target reply content.

[0181] In this embodiment, steps S703 and S704 will be repeated until agent 1 outputs exit editing or a preset number of iterations is reached. This iterative process allows agents 1 and 2 to work together to gradually improve the accuracy and rationality of the reply, thereby increasing the accuracy of the target reply content.

[0182] Step S706: Generate target reply content.

[0183] In this embodiment, the agent community's final response serves as the generative question-answering system's final answer to the user's input. This answer, after multiple rounds of collaboration and editing, mitigates issues caused by misleading retrieved information or hallucinated answers, thereby improving the accuracy of the generated target response.

[0184] In this embodiment, a large-model generative search technique based on an agent community is utilized, combined with agent community technology, to improve the quality and credibility of the target reply content of the large model (i.e., the generative search model) through multiple editing and collaboration. It can be used to establish a custom search engine and generate summary abstracts based on the large model, thereby effectively solving the information misleading and hallucination errors that may occur in the generative search scenario of the large model, thereby improving the accuracy and credibility of the reply content, providing customers with a better search experience and information summary service, thereby achieving the technical effect of improving the accuracy of the reply content and solving the technical problem of low accuracy of the reply content.

[0185] The method embodiments provided in the above embodiments of the present disclosure can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 8 is a hardware structure block diagram of a computer terminal (or mobile device) according to a method for generating interactive information in accordance with an embodiment of the present disclosure. As shown in Figure 8, the computer terminal 80 (or mobile device) may include one or more (802a, 802b, ..., 802n are used in the figure to illustrate) processors 802 (the processor 802 may include but is not limited to a microprocessor (Microcontroller Unit, referred to as MCU) or a programmable logic device (Field Programmable Gate Array, referred to as FPGA) and other processing devices), a memory 804 for storing data, and a transmission device 806 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply and / or a camera. It will be understood by those skilled in the art that the structure shown in Figure 8 is only for illustration and does not limit the structure of the above-mentioned electronic device. For example, the computer terminal 80 may also include more or fewer components than shown in FIG. 8 , or have a configuration different from that shown in FIG. 8 .

[0186] The hardware structure block diagram shown in Figure 8 can not only serve as an exemplary block diagram of the above-mentioned computer terminal 80 (or mobile device), but also as an exemplary block diagram of the above-mentioned server. In an optional embodiment, Figure 2 shows in a block diagram an embodiment of using the computer terminal 80 (or mobile device) shown in Figure 8 as a computing node in the computing environment 201.

[0187] The memory 804 can be used to store software programs and components of application software, such as the program instructions / data storage device corresponding to the data processing method in the embodiment of the present disclosure. The processor executes various functional applications and data processing by running the software programs and components stored in the memory 804, that is, realizing the above-mentioned data processing method. The memory 804 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 804 may further include a memory remotely located relative to the processor, and these remote memories may be connected to the computer terminal 80 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0188] Transmission device 806 is used to receive or transmit data via a network. A specific example of the aforementioned network may include a wireless network provided by the communications provider of computer terminal 80. In one embodiment, transmission device 806 includes a network interface controller (NIC), which can be connected to other network devices via a base station to enable communication with the Internet. In another embodiment, transmission device 806 may be a radio frequency (RF) component for wireless communication with the Internet.

[0189] The display may be, for example, a touch screen liquid crystal display (LCD), which enables a user to interact with a user interface of the computer terminal 80 (or mobile device).

[0190] It should be noted that the object information (including but not limited to object device information, object personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this disclosure are all information and data authorized by the object or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for the object to choose to authorize or refuse.

[0191] It should be noted that for the aforementioned method embodiments, for simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present disclosure is not limited by the order of the actions described, because according to the present disclosure, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and components involved are not necessarily required by the present disclosure.

[0192] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware. Based on this understanding, the technical solution of the present disclosure, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of the present disclosure.

[0193] According to an embodiment of the present disclosure, a device for generating interactive information for implementing the interactive information generation method shown in FIG3 is also provided. The device can be applied to a generative question answering system.

[0194] Figure 9 is a schematic diagram of an interactive information generation device according to an embodiment of the present disclosure. As shown in Figure 9, the interactive information generation device 900 may include: a first input component 902, a first processing component 904, a first query component 906, a first acquisition component 908 and a first replacement component 910.

[0195] The first input component 902 is configured to input query information in the operation interface of the generative question answering system.

[0196] The first processing component 904 is configured to analyze the query information using a generative search model to generate initial response content that matches the query information.

[0197] The first query component 906 is configured to query and obtain retrieval information associated with the query information, wherein the retrieval information is configured to at least represent a result retrieved based on the query information.

[0198] The first acquisition component 908 is configured to obtain associated reply content that is at least associated with the initial reply content and the retrieval information, wherein the associated reply content is generated by performing at least one round of interactive reply using at least one auxiliary generative search model and a generative search model of the generative search model.

[0199] The first replacement component 910 is configured to replace the initial response content with the associated response content to obtain target response content that matches the query information.

[0200] Here, the first input component 902, the first processing component 904, the first query component 906, the first acquisition component 908, and the first replacement component 910 correspond to steps S302 to S310 in the above embodiment. The examples and application scenarios implemented by the five components and the corresponding steps are the same, but are not limited to the contents disclosed in the above embodiment. It should be noted that the above components can be hardware components or software components stored in a memory (e.g., memory 804) and processed by one or more processors (e.g., processors 802a, 802b..., 802n). The above components can also be part of the device and can be run in the computer terminal 80 provided in Example 3.

[0201] According to an embodiment of the present disclosure, a device for generating interactive information configured to implement the interactive information generation method shown in FIG. 4 is also provided. The device can be configured to be deployed in a generative question-answering system in a search engine.

[0202] FIG10 is a schematic diagram of another device for generating interaction information according to an embodiment of the present disclosure. As shown in FIG10 , the device for generating interaction information 1000 may include: a first display component 1002 , a second display component 1004 , and a third display component 1006 .

[0203] The first display component 1002 is configured to display query information on the operation interface of the generative question answering system.

[0204] The second display component 1004 is configured to respond to the initial search instruction applied to the operation interface and display initial response content matching the query information on the operation interface, wherein the initial response content is obtained by analyzing the query information using a generative search model.

[0205] The third display component 1006 is configured to respond to a re-search instruction applied to the operation interface, and display target reply content matching the query information on the operation interface, wherein the target reply content is obtained by replacing the initial reply content with associated reply content, the retrieval information is used to at least characterize the results retrieved based on the query information, and the associated reply content is generated by performing at least one round of interactive reply using at least one auxiliary generative search model of the generative search model and the generative search model, and is associated with the initial reply content and the retrieval information.

[0206] It should be noted that the first display component 1002, the second display component 1004, and the third display component 1006 correspond to steps S402 to S406 in the above embodiment. The three components and the corresponding steps implement the same examples and application scenarios, but are not limited to the contents disclosed in the above embodiment. It should be noted that the above components can be hardware components or software components stored in a memory (e.g., memory 804) and processed by one or more processors (e.g., processors 802a, 802b..., 802n). The above components can also be part of the device and can be run in the computer terminal 80 provided in Example 3.

[0207] Figure 11 is a schematic diagram of another device for generating interactive information according to an embodiment of the present disclosure. As shown in Figure 11, the generation 1100 of interactive information may include: a second input component 1102, a second processing component 1104, a second query component 1106, a second acquisition component 1108, a second replacement component 1110 and a second output component 1112.

[0208] The second input component 1102 is configured to input query information by calling a first interface, wherein the first interface includes a first parameter, and a parameter value of the first parameter is the query information.

[0209] The second processing component 1104 is configured to analyze the query information using a generative search model to generate initial response content that matches the query information.

[0210] The second query component 1106 is configured to query and obtain retrieval information associated with the query information, wherein the retrieval information is configured to at least represent a result retrieved based on the query information.

[0211] The second acquisition component 1108 is configured to obtain associated reply content that is at least associated with the initial reply content and the retrieval information, wherein the associated reply content is generated by performing at least one round of interactive reply using at least one auxiliary generative search model and a generative search model of the generative search model.

[0212] The second replacement component 1110 is configured to replace the initial response content with the associated response content to obtain target response content that matches the query information.

[0213] The second output component 1112 is configured to output the target reply content by calling the second interface, wherein the second interface includes a second parameter, and the parameter value of the second parameter is the target reply content.

[0214] It should be noted that the second input component 1102, the second processing component 1104, the second query component 1106, the second acquisition component 1108, the second replacement component 1110, and the second output component 1112 correspond to steps S502 to S512 in the above embodiment. The six components and the corresponding steps implement the same examples and application scenarios, but are not limited to the contents disclosed in the above embodiment. It should be noted that the above components can be hardware components or software components stored in a memory (e.g., memory 804) and processed by one or more processors (e.g., processors 802a, 802b..., 802n). The above components can also be part of the device and can be run in the computer terminal 80 provided in Example 3.

[0215] In the interactive information generation device, an auxiliary generative search model is proposed. The generative search model is used to first analyze the query information. After obtaining the initial reply content, the retrieval information associated with the query information is queried, and the associated reply content generated by the interactive reply between the auxiliary generative search model and the generative search model is used to replace the initial reply content to obtain the final target reply content, thereby achieving the technical effect of improving the accuracy of the generated interactive information and solving the technical problem of low accuracy of the generated interactive information.

[0216] The embodiment of the present disclosure may provide a computer terminal, which may be any computer terminal device in a computer terminal group. Optionally, in this embodiment, the computer terminal may also be replaced by a terminal device such as a mobile terminal.

[0217] Optionally, in this embodiment, the computer terminal may be located in at least one network device among a plurality of network devices of a computer network.

[0218] In this embodiment, the above-mentioned computer terminal can execute the program code of the following steps in the method for generating interactive information: inputting query information in the operation interface of the generative question-answering system; using the generative search model to analyze the query information to generate initial reply content that matches the query information; querying to obtain retrieval information associated with the query information, wherein the retrieval information is used to at least characterize the results retrieved based on the query information; obtaining associated reply content that is at least associated with the initial reply content and the retrieval information, wherein the associated reply content is generated by performing at least one round of interactive replies using at least one auxiliary generative search model and a generative search model of the generative search model; replacing the initial reply content with the associated reply content to obtain the target reply content that matches the query information.

[0219] Optionally, Figure 12 is a structural block diagram of a computer terminal according to an embodiment of the present disclosure. As shown in Figure 12, the computer terminal A may include: one or more (only one is shown in the figure) processors 1202, a memory 1204 and a transmission device 1206.

[0220] Among them, the memory can be used to store software programs and components, such as the program instructions / components corresponding to the method and device for generating interactive information in the embodiments of the present disclosure. The processor executes various functional applications and data processing by running the software programs and components stored in the memory, that is, realizing the above-mentioned method for generating interactive information. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely located relative to the processor, and these remote memories can be connected to the computer terminal A via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0221] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: input query information in the operation interface of the generative question-answering system; use the generative search model to analyze the query information to generate initial response content that matches the query information; query to obtain retrieval information associated with the query information, wherein the retrieval information is used to at least represent the results retrieved based on the query information; obtain associated response content associated with at least the initial response content and the retrieval information, wherein the associated response content is generated by performing at least one round of interactive response using at least one auxiliary generative search model and a generative search model of the generative search model; replace the initial response content with the associated response content to obtain the target response content that matches the query information.

[0222] Optionally, the processor may also execute the program code of the following steps: according to the first prompt information, guiding the generative search model to determine the initial answer as the initial reply content when the initial answer to the query information is successfully generated; according to the first prompt information, guiding the generative search model to generate a reply framework that matches the query information when the initial answer to the query information fails to be generated, and determining the reply framework as the initial reply content, wherein the reply framework at least includes the logic for generating the initial answer.

[0223] Optionally, the above-mentioned processor can also execute the program code of the following steps: using the auxiliary generative search model to analyze the retrieval information and the initial reply content to obtain an analysis result, wherein the analysis result is used to characterize the degree of matching between the initial reply content and the retrieval information; based on the analysis result, controlling the auxiliary generative search model to generate associated reply content.

[0224] Optionally, the above-mentioned processor can also execute the program code of the following steps: if the analysis result is that the initial reply content fails to match the retrieval information, the auxiliary generative search model is controlled to adjust the initial reply content; in at least one auxiliary generative search model, when the auxiliary generative search model has a next auxiliary generative search model, the next auxiliary generative search model is used to analyze the adjusted initial reply content to obtain associated reply content.

[0225] Optionally, the above-mentioned processor can also execute the program code of the following steps: if the analysis result is that the initial reply content fails to match the retrieval information, the auxiliary generative search model is controlled to adjust the initial reply content; when the auxiliary generative search model has a next auxiliary generative search model in at least one auxiliary generative search model, the next auxiliary generative search model is used to analyze the adjusted initial reply content to obtain associated reply content.

[0226] Optionally, the processor may also execute the program code of the following steps: a first determination step, in the case where there is a next auxiliary generative search model in at least one auxiliary generative search model of the auxiliary generative search model, using the next auxiliary generative search model to determine whether the adjusted initial reply content successfully matches the query information; a second determination step, if the next auxiliary generative search model is used to determine that the adjusted initial reply content successfully matches the query information, then controlling the next auxiliary generative search model to determine the adjusted initial reply content as the associated reply content; a third determination step, if the next auxiliary generative search model is used to determine that the adjusted initial reply content fails to match the query information, then controlling the next auxiliary generative search model to adjust the adjusted initial reply content to obtain an adjustment result, and determining the next auxiliary generative search model as the auxiliary generative search model, determining the adjustment result as the adjusted initial reply content, and returning to execute the first determination step until the adjusted initial reply content successfully matches the query information, or the number of executions of the first determination step reaches a threshold number of times.

[0227] Optionally, the processor may also execute the program code of the following steps: if the analysis result is that the initial reply content fails to match the retrieval information, then determine the second prompt information associated with the retrieval information, wherein the second prompt information is used to represent the guidance information for adjusting the initial reply content; according to the second prompt information, guide the auxiliary generative search model to adjust the initial reply content.

[0228] Optionally, the processor may also execute the program code of the following steps: according to the second prompt information, guide the auxiliary generative search model, determine the target content that is allowed to be modified in the initial reply content, and adjust the target content that is allowed to be modified in the initial reply content.

[0229] Optionally, the processor may further execute program code of the following steps: if the analysis result shows that the initial reply content successfully matches the search information, the initial reply content is determined as the target reply content.

[0230] Optionally, the processor may further execute program code of the following steps: querying a database using an auxiliary generative search model to obtain retrieval information associated with the query information.

[0231] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: display the query information on the operation interface of the generative question-answering system; in response to the initial search instruction on the operation interface, display the initial reply content matching the query information on the operation interface, wherein the initial reply content is obtained by analyzing the query information using the generative search model; in response to the second search instruction on the operation interface, display the target reply content matching the query information on the operation interface, wherein the target reply content is obtained by replacing the initial reply content with the associated reply content, the retrieval information is used to at least represent the results retrieved based on the query information, and the associated reply content is generated by performing at least one round of interactive replies using at least one auxiliary generative search model and the generative search model of the generative search model, and is associated with the initial reply content and the retrieval information.

[0232] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: input the query information by calling the first interface, wherein the first interface includes a first parameter, and the parameter value of the first parameter is the query information; use the generative search model to analyze the query information to generate initial reply content that matches the query information; query to obtain retrieval information associated with the query information, wherein the retrieval information is used to at least represent the results retrieved based on the query information; obtain associated reply content associated with at least the initial reply content and the retrieval information, wherein the associated reply content is generated by performing at least one round of interactive reply using at least one auxiliary generative search model and a generative search model of the generative search model; replace the initial reply content with the associated reply content to obtain the target reply content that matches the query information; output the target reply content by calling the second interface, wherein the second interface includes a second parameter, and the parameter value of the second parameter is the target reply content.

[0233] According to the embodiment of the present disclosure, the query information is first analyzed using a generative search model to obtain the initial reply content, and then the retrieval information associated with the query information is queried and the associated reply content generated by the interactive reply between the auxiliary generative search model and the generative search model is used to replace the initial reply content to obtain the final target reply content, thereby achieving the technical effect of improving the accuracy of the generated interactive information and solving the technical problem of low accuracy of the generated interactive information.

[0234] Those skilled in the art will appreciate that the structure shown in FIG12 is merely illustrative, and that computer terminal A may also be a smartphone (e.g., an Android phone, an iOS phone, etc.), a tablet computer, a PDA, a mobile internet device (MID), a PAD, or other terminal device. FIG12 does not limit the structure of the aforementioned computer terminal A. For example, computer terminal A may include more or fewer components (e.g., a network interface, a display device, etc.) than those shown in FIG12 , or may have a configuration different from that shown in FIG12 .

[0235] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a computer-readable storage medium, which may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0236] The embodiment of the present disclosure further provides a computer-readable storage medium. Optionally, in this embodiment, the computer-readable storage medium can be used to store the program code executed by the method for generating interactive information provided in the first embodiment.

[0237] Optionally, in this embodiment, the computer-readable storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group.

[0238] Optionally, in this embodiment, the computer-readable storage medium is configured to store program codes for executing the above-mentioned execution in the computer terminal.

[0239] In the embodiment of the present disclosure, the query information is first analyzed using a generative search model to obtain initial reply content, and then the retrieval information associated with the query information is queried and the associated reply content generated by the interactive reply between the auxiliary generative search model and the generative search model is used to replace the initial reply content to obtain the final target reply content, thereby achieving the technical effect of improving the accuracy of the generated interactive information and solving the technical problem of low accuracy of the generated interactive information.

[0240] An embodiment of the present disclosure may provide an electronic device, which may include a memory and a processor.

[0241] FIG13 is a block diagram of an electronic device according to a method for generating interactive information according to an embodiment of the present disclosure. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0242] As shown in FIG13 , device 1300 includes a computing component 1301 that can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 1302 or a computer program loaded from a storage component 1308 into a random access memory (RAM) 1303. RAM 1303 can also store various programs and data required for the operation of device 1300. Computing component 1301, ROM 1302, and RAM 1303 are connected to each other via a bus 1304. An input / output (I / O) interface 1305 is also connected to bus 1304.

[0243] Various components in device 1300 are connected to I / O interface 1305, including: input component 1306, such as a keyboard, mouse, etc.; output component 1304, such as various types of displays, speakers, etc.; storage component 1308, such as a magnetic disk, optical disk, etc.; and communication component 1309, such as a network card, modem, wireless communication transceiver, etc. Communication component 1309 allows device 1300 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0244] The computing component 1301 can be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing component 1301 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing components that run machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The computing component 1301 performs the various methods and processes described above, such as the data verification method. For example, in some embodiments, the data verification method can be implemented as a computer software program that is tangibly contained in a machine-readable medium, such as the storage component 1308. In some embodiments, part or all of the computer program can be loaded and / or installed on the device 1300 via the ROM 1302 and / or the communication component 1309. When the computer program is loaded into the RAM 1303 and executed by the computing component 1301, one or more steps of the data verification method described above can be performed. Alternatively, in other embodiments, the computing component 1301 may be configured to execute the data verification method in any other appropriate manner (eg, by means of firmware).

[0245] According to an embodiment of the present disclosure, a method for generating interactive information is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0246] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system on a chip (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0247] Program code configured to implement the methods of the present disclosure may be written in any combination of one or more programming languages. Such program code may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the program code is executed by the processor or controller, the functions / operations specified in the flow charts and / or block diagrams are implemented. The program code may be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0248] In the context of the present disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium can include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0249] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or an LCD (liquid crystal display, monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0250] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0251] A computer system may include a client and a server. The client and server are generally remote from each other and typically interact through a communication network. The client-server relationship arises through computer programs running on the respective computers and having a client-server relationship with each other. The server may be a cloud server, a server in a distributed system, or a server integrated with a blockchain.

[0252] It should be noted that the serial numbers of the above-mentioned embodiments of the present disclosure are only for description and do not represent the advantages or disadvantages of the embodiments.

[0253] In the above embodiments of the present disclosure, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0254] In the several embodiments provided in this disclosure, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only exemplary. For example, the division of components is only a logical function division. In actual implementation, there may be other division methods, such as multiple components or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection of components or components can be electrical or other forms.

[0255] Components described as separate parts may or may not be physically separate, and components shown as components may or may not be physical components, that is, they may be located in one place or distributed across multiple network components. Some or all of these components may be selected based on actual needs to achieve the objectives of this embodiment.

[0256] In addition, the functional components in the various embodiments of the present disclosure may be integrated into a single processing component, each component may exist physically separately, or two or more components may be integrated into a single component. The aforementioned integrated components may be implemented in the form of hardware or software functional components.

[0257] If the integrated components are implemented in the form of software functional components and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present disclosure is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the various embodiments of the present disclosure. The aforementioned storage media include: U disk, read-only memory (ROM), random access memory (RAM), mobile hard disk, magnetic disk or optical disk, etc., various media that can store program codes.

[0258] The above is only a preferred embodiment of the present disclosure. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present disclosure. These improvements and modifications should also be regarded as within the scope of protection of the present disclosure. Industrial Applicability

[0259] The solution provided by the embodiment of the present disclosure can be applied to the process of machine learning, where query information is input into the operation interface of a generative question-answering system; the query information is analyzed using a generative search model to generate initial reply content that matches the query information; retrieval information associated with the query information is obtained by querying, wherein the retrieval information is used to at least characterize the results retrieved based on the query information; associated reply content associated with at least the initial reply content and the retrieval information is obtained, wherein the associated reply content is generated by performing at least one round of interactive replies using at least one auxiliary generative search model and a generative search model of the generative search model; the initial reply content is replaced with the associated reply content to obtain target reply content that matches the query information, thereby solving the technical problem of low accuracy of the generated interactive information.

Claims

1. A method for generating interaction information, applied to a generative question-and-answer system, the method comprising: Input query information in the operation interface of the generative question-and-answer system; Use a generative search model to analyze the query information and generate initial response content matching the query information; Query and obtain retrieval information associated with the query information, where the retrieval information is used to at least represent the result retrieved based on the query information; Obtain associated response content associated with at least the initial response content and the retrieval information, where the associated response content is generated by performing at least one round of interactive response using at least one auxiliary generative search model of the generative search model and the generative search model; Replace the initial response content with the associated response content to obtain target response content matching the query information.

2. The method according to claim 1, wherein Using the generative search model to analyze the query information and generate the initial response content matching the query information includes: Determine first prompt information associated with the query information, where the first prompt information is used to represent the guiding information for obtaining the initial response content; According to the first prompt information, guide the generative search model to generate the initial response content.

3. The method according to claim 2, wherein, According to the first prompt information, guiding the generative search model to generate the initial response content includes: According to the first prompt information, guide the generative search model to determine the initial answer as the initial response content when the generation of the initial answer to the query information is successful; According to the first prompt information, guide the generative search model to generate a response framework matching the query information and determine the response framework as the initial response content when the generation of the initial answer to the query information fails, where the response framework at least contains the logic for generating the initial answer.

4. The method according to claim 1, wherein Obtaining associated response content associated with at least the initial response content and the retrieval information includes: Use the auxiliary generative search model to analyze the retrieval information and the initial response content to obtain an analysis result, where the analysis result is used to represent the matching degree between the initial response content and the retrieval information; Based on the analysis result, control the auxiliary generative search model to generate the associated response content.

5. The method according to claim 4, wherein, Based on the analysis result, controlling the auxiliary generative search model to generate the associated response content includes: If the analysis result is that the initial response content does not match the retrieval information, then control the auxiliary generative search model to adjust the initial response content; In the at least one auxiliary generative search model, when there is a next auxiliary generative search model in the auxiliary generative search model, use the next auxiliary generative search model to analyze the adjusted initial response content to obtain the associated response content.

6. The method according to claim 5, wherein, In the at least one auxiliary generative search model, when there is a next auxiliary generative search model in the auxiliary generative search model, analyzing the adjusted initial response content by using the next auxiliary generative search model to obtain the associated response content, including: A first determination step, when there is a next auxiliary generative search model in the at least one auxiliary generative search model, using the next auxiliary generative search model to determine whether the adjusted initial response content matches the query information successfully; A second determination step, if it is determined by using the next auxiliary generative search model that the adjusted initial response content matches the query information successfully, controlling the next auxiliary generative search model to determine the adjusted initial response content as the associated response content; A third determination step, if it is determined by using the next auxiliary generative search model that the adjusted initial response content does not match the query information successfully, controlling the next auxiliary generative search model to adjust the adjusted initial response content to obtain an adjustment result, and determining the next auxiliary generative search model as the auxiliary generative search model, determining the adjustment result as the adjusted initial response content, and returning to execute the first determination step until the adjusted initial response content matches the query information successfully or the number of times of executing the first determination step reaches a threshold number of times.

7. The method according to claim 5, wherein If the analysis result is that the initial response content does not match the retrieved information, controlling the auxiliary generative search model to adjust the initial response content, including: If the analysis result is that the initial response content does not match the retrieved information, determining a second prompt message associated with the retrieved information, where the second prompt message is used to represent guiding information for adjusting the initial response content; Guiding the auxiliary generative search model to adjust the initial response content according to the second prompt message.

8. The method according to claim 7, wherein, Guiding the auxiliary generative search model to adjust the initial response content according to the second prompt message, including: Guiding the auxiliary generative search model according to the second prompt message to determine target content in the initial response content that is allowed to be modified, and adjusting the target content in the initial response content that is allowed to be modified.

9. The method according to claim 5, wherein The next auxiliary generative search model includes the generative search model.

10. The method according to claim 4, wherein The method further includes: If the analysis result is that the initial response content matches the retrieved information successfully, determining the initial response content as the target response content.

11. According to the method according to any one of claims 1 to 10, wherein Querying to obtain retrieved information associated with the query information, including: Using the auxiliary generative search model to query a database to obtain the retrieved information associated with the query information.

12. A method for generating interaction information, applied to a generative question-answering system deployed in a search engine, the method including: Displaying query information on an operation interface of the generative question-answering system; In response to a primary search instruction acting on the operation interface, initial reply content matching the query information is displayed on the operation interface, where the initial reply content is obtained by analyzing the query information using a generative search model; In response to a re-search instruction acting on the operation interface, target reply content matching the query information is displayed on the operation interface, where the target reply content is obtained by replacing the initial reply content with associated reply content, the retrieval information is used to at least represent the result retrieved based on the query information, and the associated reply content is generated by performing at least one round of interactive reply using at least one auxiliary generative search model of the generative search model and the generative search model, and is associated with the initial reply content and the retrieval information.

13. A method for generating interactive information, applied to a generative question-and-answer system, the method comprising: Inputting query information by calling a first interface, where the first interface includes a first parameter, and the parameter value of the first parameter is the query information; Analyzing the query information using a generative search model to generate initial reply content matching the query information; Querying to obtain retrieval information associated with the query information, where the retrieval information is used to at least represent the result retrieved based on the query information; Obtaining associated reply content associated with at least the initial reply content and the retrieval information, where the associated reply content is generated by performing at least one round of interactive reply using at least one auxiliary generative search model of the generative search model and the generative search model; Replacing the initial reply content with the associated reply content to obtain target reply content matching the query information; Outputting the target reply content by calling a second interface, where the second interface includes a second parameter, and the parameter value of the second parameter is the target reply content.

14. A generative question-and-answer system, comprising: A first information generation end, configured to obtain query information input in the operation interface of the generative question-and-answer system; Analyzing the query information using a generative search model to generate initial reply content matching the query information; A second information generation end, configured to perform at least one round of interactive reply using at least one auxiliary generative search model of the generative search model and the generative search model to generate associated reply content associated with at least the initial reply content and the retrieval information of the query information, where the retrieval information is used to at least represent the result retrieved based on the query information; wherein the first information generation end is configured to replace the initial reply content with the associated reply content and output target reply content matching the query information.

15. An electronic device, comprising: A memory and a processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, and when the computer-executable instructions are executed by the processor, the steps of any one of the methods from 1 to 13 are implemented.

16. A computer-readable storage medium, the computer-readable storage medium including a stored program, wherein, When the program is run by a processor, control the device where the computer-readable storage medium is located to execute the method according to any one of claims 1 to 13.

Images