Interaction information processing method and system, electronic equipment and storage medium

Abstract

The invention discloses an interaction information processing method and system, electronic equipment and a storage medium, and relates to the field of natural language processing, machine learning and artificial intelligence. The method comprises the steps that inquiry information to be replied is acquired, and the inquiry information is used for inquiring reply information associated with a product; historical dialogue information associated with the inquiry information is obtained, the historical dialogue information comprises at least one piece of historical inquiry information and historical reply information matched with the historical inquiry information, and the historical inquiry information and the historical reply information are generated by conducting at least one round of interactive reply on the product; the quality index of the historical reply information is higher than a quality index threshold; generating prompt information based on the inquiry information and the historical dialogue information; and utilizing the prompt information to guide the information processing model to generate reply information matched with the inquiry information. The technical problem that the processing effect of the interaction information is poor is solved.

Description

Technical Field

[0001] The present application relates to the fields of natural language processing, machine learning and artificial intelligence, and more specifically, to a method, system, electronic device and storage medium for processing interactive information. Background Art

[0002] Currently, for product consultation scenarios, it is necessary to generate reply information that matches the inquiry information based on the user's actual inquiry information.

[0003] In the related art, the historical conversation information is usually extracted by performing Cartesian product on the historical query information and the historical reply information, and then the product knowledge is filtered out offline through the semantic matching model. The query information to be replied is retrieved online by similarity calculation and other retrieval methods, and the reply information is returned. However, the integrity and quality of the reply information returned by this method are poor, resulting in the problem of poor processing effect of interactive information.

[0004] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiments of the present application provide a method, system, electronic device and storage medium for processing interactive information, so as to at least solve the technical problem of poor processing effect of interactive information.

[0006] According to one aspect of an embodiment of the present application, a method for processing interactive information is provided. The method may include: obtaining inquiry information to be answered, wherein the inquiry information is used to inquire about answer information associated with a product; obtaining historical conversation information associated with the inquiry information, wherein the historical conversation information includes at least one historical inquiry information and historical answer information matching the historical inquiry information, the historical inquiry information and the historical answer information are generated for at least one round of interactive answering for the product, and the quality index of the historical answer information is higher than the quality index threshold; based on the inquiry information and the historical conversation information, generating prompt information; using the prompt information, guiding an information processing model to generate answer information matching the inquiry information, wherein the information processing model is trained using prompt information samples and answer information samples associated with the product, and the prompt information samples include inquiry information samples matching the answer information samples.

[0007] According to another aspect of the embodiment of the present application, a method for determining a model is provided. The method may include: obtaining a prompt information sample in an e-commerce scenario, wherein the prompt information sample is used to prompt the generation of a reply information sample that matches the inquiry information sample in the prompt information sample; using the prompt information sample and the reply information sample to train an information processing model, wherein the information processing model is guided by the prompt information to generate reply information that matches the inquiry information, the prompt information is generated based on the inquiry information and the historical dialogue information associated with the inquiry information, the inquiry information is used to inquire about the reply information associated with the product in the e-commerce scenario, the historical dialogue information includes at least one historical inquiry information, and the historical reply information that matches the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive replies to the product, and the quality index of the historical reply information is higher than the quality index threshold.

[0008] According to another aspect of the embodiment of the present application, a method for processing interactive information is provided. The method may include: identifying inquiry information to be answered on an e-commerce platform, wherein the inquiry information is used to inquire about answer information associated with a product on the e-commerce platform; obtaining historical conversation information associated with the inquiry information, wherein the historical conversation information includes at least one historical inquiry information on the e-commerce platform, and historical answer information matching the historical inquiry information, the historical inquiry information and the historical answer information are generated for at least one round of interactive answering for the product, and the quality index of the historical answer information is higher than the quality index threshold; generating prompt information based on the inquiry information and the historical conversation information; using the prompt information, guiding an information processing model to generate answer information on the e-commerce platform, wherein the answer information matches the inquiry information, wherein the information processing model is trained using prompt information samples and answer information samples generated on the e-commerce platform, and the prompt information samples include inquiry information samples matching the answer information samples.

[0009] According to another aspect of the embodiment of the present application, a method for processing interactive information is provided. The method may include: in response to an inquiry instruction acting on an operation interface, displaying inquiry information to be answered on the operation interface, wherein the inquiry information is used to inquire about answer information associated with a product; in response to an information answer instruction acting on the operation interface, displaying historical dialogue information associated with the inquiry information on the operation interface, wherein the historical dialogue information includes at least one historical inquiry information and historical answer information matching the historical inquiry information, the historical inquiry information and the historical answer information are generated for at least one round of interactive answering for the product, and the quality index of the historical answer information is higher than the quality index threshold; displaying prompt information on the operation interface, wherein the prompt information is generated based on the inquiry information and the historical dialogue information; displaying answer information matching the inquiry information on the operation interface, wherein the answer information is generated by using the prompt information to guide an information processing model, the information processing model is trained using prompt information samples and answer information samples associated with the product, and the prompt information samples include inquiry information samples matching the answer information samples.

[0010] According to another aspect of the embodiment of the present application, a method for processing interactive information is provided. The method may include: obtaining inquiry information to be answered by calling a first interface, wherein the first interface includes a first parameter, the parameter value of the first parameter includes inquiry information, and the inquiry information is used to inquire about the reply information associated with the product; obtaining historical dialogue information associated with the inquiry information, wherein the historical dialogue information includes at least one historical inquiry information and historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive reply for the product, and the quality index of the historical reply information is higher than the quality index threshold; based on the inquiry information and the historical dialogue information, generating prompt information; using the prompt information, guiding the information processing model to generate reply information matching the inquiry information, wherein the information processing model is trained using the prompt information sample and the reply information sample associated with the product, and the prompt information sample includes the inquiry information sample matching the reply information sample; outputting the reply information by calling a second interface, wherein the second interface includes a second parameter, and the parameter value of the second parameter includes the reply information.

[0011] According to another aspect of an embodiment of the present application, an electronic device is also provided. The electronic device 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, any one of the above-mentioned methods is implemented.

[0012] According to another aspect of an embodiment of the present application, a processor is further provided, and the processor 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 application, a computer-readable storage medium is further provided, the computer-readable storage medium including a stored program, wherein when the program is running, the device where the storage medium is located is controlled to execute any of the above methods.

[0014] According to another aspect of an embodiment of the present application, a computer program product is also provided, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, any one of the above methods is implemented.

[0015] In an embodiment of the present application, query information to be answered is obtained, wherein the query information is used to inquire about answer information associated with a product; historical conversation information associated with the query information is obtained, wherein the historical conversation information includes at least one historical query information and historical answer information matching the historical query information, the historical query information and the historical answer information are generated for at least one round of interactive answering for the product, and the quality index of the historical answer information is higher than the quality index threshold; prompt information is generated based on the query information and the historical conversation information; using the prompt information, an information processing model is guided to generate answer information matching the query information, wherein the information processing model is trained using prompt information samples and answer information samples associated with the product, and the prompt information samples include query information samples matching the answer information samples. That is to say, the embodiment of the present application obtains high-quality historical conversation information associated with the query information to be answered, and uses the high-quality historical conversation information to optimize and supplement the answer information matching the query information, thereby achieving the purpose of making the generated answer information closer to the user's real expected answer, thereby achieving the improvement of the processing effect of the interactive information, and solving the technical problem of poor processing effect of the interactive information.

[0016] It is easy to notice that the above general description and the following detailed description are only for the purpose of exemplifying and explaining the present application, and do not constitute a limitation of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0018] Figure 1 is a schematic diagram of an application scenario of a method for processing interaction information according to an embodiment of the present application;

[0019] Figure 2 is a flowchart of a method for processing interaction information according to an embodiment of the present application;

[0020] Figure 3 is a flow chart of a method for determining a model according to an embodiment of the present application;

[0021] Figure 4 is a flowchart of another method for processing interaction information according to an embodiment of the present application;

[0022] Figure 5 is a flowchart of another method for processing interaction information according to an embodiment of the present application;

[0023] Figure 6 is a flowchart of another method for processing interaction information according to an embodiment of the present application;

[0024] Figure 7 is a schematic diagram of guiding a large conversation understanding model to generate intelligent answers according to an embodiment of the present application;

[0025] Figure 8 is a schematic diagram of recalling historical conversation records according to an embodiment of the present application;

[0026] Fig. 9 is a structural block diagram of a computing environment of a method for processing interactive information according to an embodiment of the present application;

[0027] Fig.10 is a schematic diagram of a device for processing interactive information according to an embodiment of the present application;

[0028] Fig.11 is a schematic diagram of a device for determining a model according to an embodiment of the present application;

[0029] Fig.12 is a schematic diagram of another interactive information processing device according to an embodiment of the present application;

[0030] Fig.13 is a schematic diagram of another interactive information processing device according to an embodiment of the present application;

[0031] Fig.14 is a schematic diagram of another interactive information processing device according to an embodiment of the present application;

[0032] Fig.15 is a structural block diagram of a computer terminal according to an embodiment of the present application;

[0033] Fig.16 is a block diagram of an electronic device according to a method for processing interactive information in an embodiment of the present application;

[0034] Fig.17 It is a hardware structure block diagram of a computer terminal (or mobile device) for implementing a method for processing interactive information according to an embodiment of the present application. DETAILED DESCRIPTION

[0035] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0036] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application 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 data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0037] First, some nouns or terms that appear in the description of the embodiments of the present application are subject to the following explanations:

[0038] Supervised fine-tuning (SFT) is a deep learning strategy that can be used for pre-trained large models. In supervised fine-tuning, the model parameters are initialized by pre-training on a large-scale dataset, and supervised fine-tuning is performed on a specific task to further optimize the model parameters to adapt to the task. It can be used to speed up the model convergence and improve the performance and generalization ability of the model.

[0039] Retrieval Augmented Generation (RAG), a method for optimizing the output of large language models by enabling them to reference authoritative knowledge bases outside of the training data before generating a response;

[0040] Chain-of-thought (CoT), an improved prompting technique, refers to a series of logically related thinking steps that form a complete thinking process. In the chain of thought, each thinking node has a direct logical connection with the previous node, forming an organic thinking structure that can be used to determine the connection and logic between problems, thereby forming a more complete and reasonable viewpoint or conclusion.

[0041] The above-mentioned interactive information processing method provided in the embodiment of the present application can be applied to the following Figure 1 The application scenarios shown are not limited to these. Figure 1 is a schematic diagram of an application scenario of a method for processing interactive information according to an embodiment of the present application. Figure 1 In the application scenario shown, the server 10 can be a cloud. The server 10 can be connected to one or more client devices 20 via a local area network connection, a wide area network connection, an Internet connection, or other types of data networks. The client devices 20 here may include but are not limited to: smart phones, tablet computers, laptops, PDAs, personal computers, smart home devices, vehicle-mounted devices, etc. The client devices together constitute the client relative to the server. An operating interface for displaying inquiry information to be answered can be deployed on the graphical user interface on the client device. The client device 20 can interact with the user through the graphical user interface to implement the method for processing interactive information provided in the embodiment of the present application.

[0042] In the embodiment of the present application, the system composed of the client device 20 and the server 10 can perform the following steps: if there is a need to obtain the inquiry information to be answered, the corresponding operation can be performed in the operation interface on the client device 20 to obtain the inquiry information to be answered. The client device can obtain the inquiry information and send it to the server through the network. After receiving the inquiry information, the server can perform the following steps: step S102, obtain the inquiry information to be answered, wherein the inquiry information is used to inquire about the reply information associated with the product; step S104, obtain the historical dialogue information associated with the inquiry information, wherein the historical dialogue information includes at least one historical inquiry information and the historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive reply for the product, and the quality index of the historical reply information is higher than the quality index threshold; step S106, based on the inquiry information and the historical dialogue information, generate prompt information; step S108, use the prompt information to guide the information processing model to generate the reply information matching the inquiry information, wherein the information processing model is trained using the prompt information sample and the reply information sample associated with the product, and the prompt information sample includes the inquiry information sample matching the reply information sample.

[0043] Under the above operating environment, this application provides Figure 2 The method for processing the interactive information shown. Figure 2 is a flowchart of a method for processing interactive information according to an embodiment of the present application. Figure 2 As shown, the method may include the following steps:

[0044] Step S202, obtaining inquiry information to be answered.

[0045] In the technical solution provided in the above step S202 of the present application, inquiry information to be answered can be obtained. Among them, the inquiry information can be used to inquire about the reply information associated with the product. Optionally, in an e-commerce scenario, the product can be a commodity to be traded, and the inquiry information can be used to inquire about the reply information associated with the commodity, for example, the inquiry information can be used to inquire about the reply information such as the quality, attributes and performance of the commodity. The reply information can be information that responds to the inquiry information, for example, it can be high-quality reply information corresponding to the inquiry information generated by intelligent customer service, which can be called intelligent reply information or answer information.

[0046] For example, in an e-commerce scenario, the inquiry information may be real question information asked by the user, such as "What is the material of the product?", "What is the size of the product?", "How long does it take to ship after purchasing the product?", "Who is the product suitable for?" and other real questions asked by the user.

[0047] It should be noted that the above-mentioned commodities in the e-commerce scenario and the inquiry information associated with the commodities are only examples, and do not specifically limit the application scenario and the types of products and inquiry information in the application scenario.

[0048] Step S204: Acquire historical conversation information associated with the inquiry information.

[0049] In the technical solution provided in the above step S204 of the present application, after obtaining the inquiry information to be answered, historical conversation information associated with the inquiry information to be answered can be obtained. The historical conversation information may include at least one historical inquiry information and historical answer information matching the historical inquiry information.

[0050] In this embodiment, the above historical inquiry information may be information associated with the product inquired by the current user or other users in the historical period. The historical reply information may be information for replying to the historical inquiry information, for example, the historical reply information may be information for replying to the historical inquiry information related to the product by the manual customer service in the e-commerce scenario. The historical dialogue information may be called historical session information, and the historical session information may be the manual customer service historical session information.

[0051] It should be noted that the above historical inquiry information and historical reply information are generated for at least one round of interactive reply for the product, that is, the historical inquiry information and the historical reply information can constitute one or more groups of dialogue information. The quality index of the historical reply information is higher than the quality index threshold. Among them, the quality index threshold can be an index threshold set according to actual conditions, and the quality index threshold can be adjusted according to actual needs.

[0052] In this embodiment, after obtaining the inquiry information to be answered, historical conversation information associated with the inquiry information to be answered can be obtained. The historical conversation information may include one or more groups of historical inquiry information and historical reply information. The historical reply information can be high-quality information of the manual customer service's reply to the historical inquiry information. This is only for illustration, and no specific limitation is imposed on the method of generating the historical reply information.

[0053] Step S206: Generate prompt information based on the query information and the historical dialogue information.

[0054] In the technical solution provided in the above step S206 of the present application, after obtaining the inquiry information to be replied and the historical conversation information associated with the inquiry information, prompt information can be generated based on the obtained inquiry information and historical conversation information.

[0055] In this embodiment, after obtaining the inquiry information to be replied and the historical dialogue information associated with the inquiry information, the obtained inquiry information and the historical dialogue information associated with the inquiry information can be combined to obtain prompt information. Optionally, this embodiment combines the inquiry information and the historical dialogue information associated with the inquiry information according to an information combination format that matches the information processing model to obtain prompt information, so that the prompt information can be input into the information processing model.

[0056] Step S208, using the prompt information, guides the information processing model to generate reply information that matches the inquiry information. In the technical solution provided in the above step S208 of the present application, after the prompt information is generated based on the inquiry information and the historical conversation information, the prompt information is input into the information processing model, and the prompt information is used to guide the information processing model to generate reply information that matches the inquiry information. Among them, the information processing model can be a model trained using prompt information samples and reply information samples associated with the product, which can be called a conversation understanding model. The prompt information sample can at least include an inquiry information sample that matches the reply information sample, which can be called a prompt word sample. The inquiry information sample can at least include historical inquiry information, and the reply information sample can at least include historical reply information, and the reply information sample can be called an answer sample.

[0057] Optionally, the samples associated with the above-mentioned product may be generated by inquiring about the product, and the samples associated with the product may be samples in the scenario where the product is located, wherein the scenario where the product is located may be an e-commerce scenario, an education scenario, a medical scenario, a government scenario, etc. This is only an example and does not specifically limit the scenario where the product is located. For example, in an e-commerce scenario, prompt information samples and reply information samples in the e-commerce scenario may be obtained, and the information processing model may be trained using the obtained prompt information samples and reply information samples.

[0058] In this embodiment, after obtaining the inquiry information to be replied, the obtained inquiry information can be used to obtain high-quality historical dialogue information associated with the inquiry information. After obtaining the high-quality historical dialogue information associated with the inquiry information, the obtained inquiry information and the high-quality historical dialogue information can be used as inputs of the information processing model to guide the information processing model to generate reply information that matches the inquiry information. Optionally, the reply information is the reply information after optimizing and supplementing the intelligent answer information using the historical conversation information of the manual customer service. The reply information can be closer to the user's actual expected answer, thereby improving the processing effect of the interactive information.

[0059] Through the above steps S202 to S208 of the present application, the inquiry information to be answered is obtained, wherein the inquiry information is used to inquire about the answer information associated with the product; the historical dialogue information associated with the inquiry information is obtained, wherein the historical dialogue information includes at least one historical inquiry information and the historical answer information matching the historical inquiry information, the historical inquiry information and the historical answer information are generated for at least one round of interactive answering for the product, and the quality index of the historical answer information is higher than the quality index threshold; based on the inquiry information and the historical dialogue information, prompt information is generated; using the prompt information, the information processing model is guided to generate the answer information matching the inquiry information, wherein the information processing model is trained using the prompt information sample and the answer information sample associated with the product, and the prompt information sample includes the inquiry information sample matching the answer information sample. That is to say, the embodiment of the present application obtains high-quality historical dialogue information associated with the inquiry information to be answered, and uses the high-quality historical dialogue information to optimize and supplement the answer information matching the inquiry information, thereby achieving the purpose of making the generated answer information closer to the user's real expected answer, thereby achieving the improvement of the processing effect of the interactive information, and solving the technical problem of poor processing effect of the interactive information.

[0060] The above method of this embodiment is further introduced below.

[0061] As an optional implementation, step S204, obtaining historical conversation information associated with the inquiry information, includes: searching in an external information repository of the information processing model according to the inquiry information to obtain historical conversation information, wherein the external information repository includes conversation information other than prompt information samples and reply information samples, and the conversation information includes different types of knowledge about the product.

[0062] In this embodiment, after obtaining the inquiry information to be replied, the external information base of the information processing model can be searched according to the obtained inquiry information to obtain the historical dialogue information. The external information base can at least include dialogue information other than the prompt information sample and the reply information sample, and the dialogue information can at least include different types of product knowledge. The external information base can be an authoritative knowledge base other than the training data such as the prompt information sample and the reply information sample, which can be called a knowledge base.

[0063] Optionally, after obtaining the inquiry information to be replied, according to the obtained inquiry information, a search can be performed in the authoritative knowledge base of the information processing model, and the output information of the RAG optimized information processing model can be generated by retrieval enhancement, so that the information processing model can refer to the authoritative knowledge base other than the training data such as the prompt information sample and the reply information sample before generating the reply information matching the inquiry information, thereby obtaining the historical conversation information associated with the inquiry information. That is, this embodiment achieves the purpose of recalling the historical conversation information of the manual customer service through RAG.

[0064] Optionally, after obtaining the historical conversation information associated with the inquiry information, the historical conversation information can be preprocessed through intent recognition, invalid speech recognition and clustering algorithms, for example, the historical conversation information can be screened, cleaned and deduplicated, etc., without specific limitations here.

[0065] As an optional implementation, according to the query information, a search is performed in an external information repository of the information processing model to obtain historical conversation information, including: vectorizing the query information; according to the vectorized query information, a search is performed in the external information repository to obtain a search result, and the historical conversation information corresponding to the search result is determined, wherein the search result is used to represent the vectorized historical conversation information.

[0066] In this embodiment, after obtaining the inquiry information to be replied, the inquiry information can be vectorized to obtain the vectorized inquiry information. Further, according to the obtained vectorized inquiry information, a search can be performed in an external information database to obtain a search result, and the historical conversation information corresponding to the search result can be determined. The search result can be used to represent the vectorized historical conversation information.

[0067] Optionally, after obtaining the inquiry information to be replied, the inquiry information can be vectorized using a question-answering model (e.g., StructBERTFAQ question-answering model) to obtain the vectorized inquiry information. Further, according to the obtained vectorized inquiry information, a search can be performed in an authoritative knowledge base to obtain a search result, which can be used to represent the vectorized historical conversation information, and the StructBERT FAQ question-answering model can be used to vectorize the historical conversation information. That is, this embodiment uses the StructBERT FAQ question-answering model to vectorize the real question information asked by the user and the historical conversation information in the knowledge base and then recall them.

[0068] As an optional implementation, a search is performed in an external information repository according to the vectorized query information to obtain a search result, including: a search is performed in an external information repository according to the vectorized query information to obtain vectorized historical query information associated with the product; a search is performed in an external information repository according to the vectorized historical query information to obtain a search result including the vectorized historical query information and the vectorized historical reply information.

[0069] In this embodiment, after the query information is vectorized to obtain the vectorized query information, a search can be performed in the external information database according to the vectorized query information to obtain the vectorized historical query information associated with the product. After the vectorized historical query information associated with the product is obtained, a search can be performed in the external information database according to the vectorized historical query information to obtain a search result including the vectorized historical query information and the vectorized historical reply information.

[0070] Optionally, after the query information is vectorized using the StructBERT FAQ question-answering model to obtain the vectorized query information, the vectorized query information can be searched in the authoritative knowledge base through RAG to obtain the vectorized historical query information associated with the product, for example, the historical consultation questions of the product. Further, according to the vectorized historical query information, the authoritative knowledge base can be searched through RAG to obtain retrieval results including the vectorized historical query information and the vectorized historical reply information. Among them, the StructBERT FAQ question-answering model can be used to vectorize the historical query information and the historical reply information to obtain the vectorized historical query information and the vectorized historical reply information.

[0071] As an optional implementation, a search is performed in an external information repository according to the vectorized historical query information to obtain a search result including the vectorized historical query information and the vectorized historical reply information, including: using the vectorized historical query information as index information, indexing in the external information repository the search result including the vectorized historical query information and the vectorized historical reply information.

[0072] In this embodiment, after obtaining the vectorized historical query information associated with the product, the vectorized historical query information is used as index information to index retrieval results including the vectorized historical query information and the vectorized historical reply information in an external information repository.

[0073] Optionally, after the historical conversation information is vectorized using the StructBERT FAQ question-answering model to obtain the vectorized historical conversation information, multiple historical inquiry information related to the historical conversation information can be extracted from the vectorized historical conversation information, that is, product consultation questions can be extracted as index information of the historical conversation information. Further, according to the index information, a search result including the vectorized historical inquiry information and the vectorized historical answer information can be indexed in the authoritative knowledge base.

[0074] As an optional implementation, step S206, generating prompt information based on the query information and the historical conversation information, includes: performing semantic recognition on the context in the historical conversation information to obtain semantic recognition information; and combining the query information and the semantic recognition information to obtain prompt information.

[0075] In this embodiment, after acquiring the historical conversation information associated with the inquiry information, semantic recognition may be performed on the context in the historical conversation information to obtain semantic recognition information. After semantic recognition is performed on the context in the historical conversation information to obtain semantic recognition information, the inquiry information and the semantic recognition information may be combined to obtain prompt information. The semantic recognition information may be information obtained by semantic recognition of the context in the historical conversation information.

[0076] Optionally, after obtaining the historical conversation information associated with the inquiry information, semantic recognition can be performed on the context in the historical conversation information to obtain semantic recognition information. That is, the semantic recognition information in this embodiment has the ability to understand the context of the historical conversation information, for example, in an e-commerce scenario, it has the ability to understand the conversation in the e-commerce scenario. After obtaining the semantic recognition information, the inquiry information and the semantic recognition information can be combined to obtain prompt information.

[0077] In the related art, the knowledge of the historical conversation information is extracted by performing a Cartesian product on the real question information of the consumer or user and the answer information of the customer service in a historical conversation information, and then the product knowledge is filtered out offline through a semantic matching model (for example, a QA semantic matching model) and put into the product knowledge base. The extracted answer information is returned online through similarity calculation and other retrieval methods based on the real question information of the user. However, this method has no contextual understanding ability for the historical conversation information, and the accuracy of the QA knowledge extracted by QA text matching is low. At the same time, the generated answer information is only the single-sentence answer information that has appeared in the historical conversation information, and there is no ability to summarize the answer information of multiple rounds of customer service based on historical conversations, resulting in poor integrity and quality of the generated answer information. In addition, the offline link knowledge production method cannot utilize the historical conversation information in real time, and relies on the QA knowledge retrieval capability, resulting in low effectiveness and utilization of the generated answer information, resulting in poor processing effect of interactive information.

[0078] However, in the embodiment of the present application, in order to solve the problem of poor processing effect of interactive information in the above-mentioned related technologies, the relevant historical conversation information is recalled according to the user's real question information, and high-quality answer information is generated based on the big model, and the answer is given to the consumer in real time, thus realizing the closed loop between manual customer service and intelligence. That is, this embodiment realizes the real-time optimization and supplementation of the answer information on the intelligent side by the manual customer service historical conversation information through the big model of conversation understanding, greatly reducing the cost of manual customer service, and the answer information generated based on the real manual customer service historical conversation information is closer to the real expected answer of the consumer, significantly improving the pre-sales satisfaction, thereby realizing the improvement of the processing effect of interactive information.

[0079] As an optional implementation, prompt information is used to guide the information processing model to generate response information that matches the query information, including: controlling the information processing model to infer the prompt information according to information reasoning rules to obtain the response information, wherein the information reasoning rules are used to represent the logical relationship of the information processing model's reasoning of the prompt information.

[0080] In this embodiment, after the query information and the semantic recognition information are combined to obtain the prompt information, the information processing model can be controlled to infer the prompt information according to the information inference rule to obtain the reply information. The information inference rule can be used to represent the logical relationship of the information processing model in inferring the prompt information, so that the information processing model can realize the inference of complex tasks.

[0081] Optionally, the data format required for supervised fine-tuning SFT is (prompt information, reply information). In order to improve the accuracy of reply information generated by the conversation understanding large model, after combining the query information and the semantic recognition information to obtain the prompt information, the chain of thought CoT technology can be used in the prompt. Since CoT refers to a series of logically related thinking steps that can form a complete thinking process, the reply information can include two parts: the intermediate reasoning result and the answer, so that the conversation understanding large model can better realize the reasoning of complex tasks.

[0082] Optionally, in an online link, this embodiment recalls multiple segments of historical conversation information associated with the real question information of the current user through RAG, and assembles them into prompts as inquiries for the conversation understanding large model.

[0083] As an optional implementation, the information processing model is controlled to infer the prompt information according to the information reasoning rules to obtain the reply information, including: when the prompt information indicates that the reply information to be generated includes the answer to the inquiry information, the information processing model is controlled to infer the prompt information according to the information reasoning rules, and output the reply information including the answer to the inquiry information; when the prompt information indicates that the reply information to be generated includes the intermediate reasoning result and the answer to the inquiry information, the information processing model is controlled to infer the prompt information according to the information reasoning rules, and output the reply information including the intermediate reasoning result and the answer, wherein the intermediate reasoning result is used to enable the information processing model to infer the answer.

[0084] In this embodiment, after the query information and the semantic recognition information are combined to obtain the prompt information, when the prompt information indicates that the reply information to be generated includes the answer to the query information, the information processing model can be controlled to infer the prompt information according to the information reasoning rules, and the reply information including the answer to the query information can be output. When the prompt information indicates that the reply information to be generated includes the intermediate reasoning result and the answer to the query information, the information processing model can be controlled to infer the prompt information according to the information reasoning rules, and the reply information including the intermediate reasoning result and the answer can be output. Among them, the intermediate reasoning result can be used to enable the information processing model to infer the answer, which can be called intermediate thinking.

[0085] Optionally, when the prompt information indicates that the reply information to be generated includes the intermediate reasoning result and the answer of the inquiry information, that is, the intermediate thinking is required to be output, then the type of training data is (prompt information-required to output intermediate thinking, reply information). Therefore, after the control information processing model is inferred on the prompt information according to the information reasoning rule, the reply information including the intermediate reasoning result and the answer can be output.

[0086] Optionally, in the process of generating the reply information, if intermediate thoughts are generated, a large delay will be caused. Taking into account the performance requirements of the online model, this embodiment simultaneously constructs a part of the prompt information and the corresponding answer information that do not require the generation of intermediate thoughts. In the case where the prompt information indicates that the reply information to be generated includes the answer to the inquiry information, that is, it is not required to output the intermediate thoughts, the type of training data is (prompt information-does not require the output of intermediate thoughts, reply information). Therefore, after the control information processing model is inferred for the prompt information according to the information inference rules, the reply information including the answer to the inquiry information can be output.

[0087] It should be noted that in order to ensure the comprehensiveness of the training data, this embodiment randomly extracts a certain number of users' real question information and manual customer service conversation information associated with the real question information for different question intentions, and uses Generative Pre-trained Transformer 4 (GPT4) to query the optimized prompt to generate answer information and intermediate thinking corresponding to the real question information.

[0088] As an optional implementation, the method further includes: in the case where the answer fails to match the query information, the information processing model can be controlled to output response information that does not include the answer.

[0089] In this embodiment, in order to maintain the ability of the large conversation understanding model to refuse to answer, a certain proportion of no-answer replies needs to be mixed into the training data. That is, when the answer fails to match the query information, the information processing model can be controlled to output reply information that does not include the answer.

[0090] In an embodiment of the present application, high-quality historical conversation information associated with the inquiry information to be answered is obtained, and the reply information matching the inquiry information is optimized and supplemented using the high-quality historical conversation information, thereby achieving the purpose of making the generated reply information closer to the user's actual expected answer, thereby achieving the improvement of the processing effect of the interactive information and solving the technical problem of poor processing effect of the interactive information.

[0091] In view of the training of the above information processing model, the embodiment of the present application also provides a method for determining a model. Figure 3 is a flow chart of a method for determining a model according to an embodiment of the present application, such as Figure 3 As shown, the method may include the following steps:

[0092] Step S302, obtaining prompt information samples in the e-commerce scenario.

[0093] In the technical solution provided in the above step S302 of the present application, a prompt information sample in an e-commerce scenario can be obtained. The prompt information sample can be used to prompt the generation of a reply information sample that matches the inquiry information sample in the prompt information sample. The inquiry information sample can at least include historical inquiry information, and the reply information sample can at least include historical reply information. The reply information sample can be called an answer sample.

[0094] Step S304: train an information processing model using the prompt information samples and the reply information samples.

[0095] In the technical solution provided in the above step S304 of the present application, after obtaining the prompt information samples in the e-commerce scenario, the prompt information samples and the reply information samples can be used to train an information processing model. The information processing model can be guided by the prompt information to generate reply information matching the inquiry information, and the information processing model can be called a conversation understanding model.

[0096] Optionally, the inquiry information can be used to inquire about reply information associated with products in the e-commerce scenario. For example, the inquiry information can be used to inquire about information such as the quality, attributes, and performance of the product. The inquiry information can be real question information asked by the user, for example, the inquiry information can be "What is the material of the product?", "What is the size of the product?", "How long after purchasing the product can it be shipped?", "Who is the product suitable for?", etc. The reply information can be information that responds to the inquiry information, for example, it can be high-quality reply information corresponding to the inquiry information generated by the intelligent customer service, which can be called intelligent reply information or answer information.

[0097] It should be noted that the above-mentioned commodities in the e-commerce scenario and the inquiry information associated with the commodities are only examples, and do not specifically limit the application scenario and the types of products and inquiry information in the application scenario.

[0098] Optionally, the historical dialogue information may include at least one historical inquiry information and historical reply information matching the historical inquiry information. The historical inquiry information may be information associated with the product inquired by the current user or other users in the historical period. The historical reply information may be information that replies to the historical inquiry information, for example, the historical reply information may be information that a human customer service representative in an e-commerce scenario replies to historical inquiry information related to the product. The historical dialogue information may be referred to as historical session information, and the historical session information may be human customer service historical session information.

[0099] It should be noted that the above historical inquiry information and historical reply information are generated for at least one round of interactive reply for the product, that is, the historical inquiry information and the historical reply information can constitute one or more groups of dialogue information. The quality index of the historical reply information is higher than the quality index threshold. Among them, the quality index threshold can be an index threshold set according to actual conditions, and the quality index threshold can be adjusted according to actual needs.

[0100] Optionally, the prompt information may be generated based on the query information and the historical dialogue information associated with the query information. The query information and the historical dialogue information associated with the query information may be combined to obtain the prompt information. Optionally, this embodiment combines the query information and the historical dialogue information associated with the query information in an information combination format that matches the information processing model to obtain the prompt information, so that the prompt information is allowed to be input into the information processing model.

[0101] Through the above-mentioned steps S302 and S304 of the present application, prompt information samples in the e-commerce scenario are obtained, wherein the prompt information samples are used to prompt the generation of reply information samples that match the inquiry information samples in the prompt information samples; the prompt information samples and the reply information samples are used to train an information processing model, wherein the information processing model is guided by the prompt information to generate reply information that matches the inquiry information, the prompt information is generated based on the inquiry information and the historical dialogue information associated with the inquiry information, the inquiry information is used to inquire about reply information associated with the product in the e-commerce scenario, the historical dialogue information includes at least one historical inquiry information, and historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive replies to the product, and the quality index of the historical reply information is higher than the quality index threshold, thereby achieving the purpose of making the generated reply information closer to the user's real expected answer, thereby achieving the improvement of the processing effect of the interactive information and solving the technical problem of poor processing effect of the interactive information.

[0102] The above method of this embodiment is further introduced below.

[0103] As an optional implementation, step S304, using prompt information samples and reply information samples to train an information processing model, includes: determining a pre-trained model in an e-commerce scenario, wherein the pre-trained model is used to represent a large language model that allows training in an e-commerce scenario; adding a sub-model to the pre-trained model, wherein the sub-model is used to represent a matrix with an order less than an order threshold; using prompt information samples and reply information samples to train the sub-model; and determining the pre-trained model and the trained sub-model as the information processing model.

[0104] In this embodiment, a pre-trained model in an e-commerce scenario can be determined. After determining the pre-trained model in the e-commerce scenario, a sub-model can be added to the determined pre-trained model. After obtaining the prompt information sample in the e-commerce scenario, the sub-model can be trained using the obtained prompt information sample and reply information sample. The pre-trained model and the trained sub-model can be further determined as an information processing model. Among them, the pre-trained model can be used to represent a large language model that allows training in an e-commerce scenario, which can be called a base model (for example, qwen1.5 7B), a pre-trained matrix, and can be represented by W0. The sub-model can be used to represent a matrix whose order is less than an order threshold, which can be called a low-order matrix and can be represented by ΔW. The order threshold can be a threshold set according to actual conditions, and the order threshold can be adjusted in real time according to actual needs.

[0105] Optionally, based on the user's question information and the historical conversation information of the manual customer service related to the question information, the large model can generate answer information corresponding to the question information. First, the base model needs to have a strong ability to understand e-commerce scene dialogues and be able to summarize accurate and high-quality answer information from the dialogue. Secondly, the base model needs to have a low hallucination rate, and should refuse to answer questions that cannot be answered accurately based on the content of the dialogue. Finally, the base model needs a certain ability to follow instructions and can give answer information in a specified format. In summary, it can be determined that the base model is qwen1.5 7B.

[0106] Optionally, after determining the pre-trained model, Local Outlier Rejection and Acceptance (LORA) is selected as a fine-tuning method. LORA can be used to add sub-models, i.e., add low-order matrices, and learn new task-specific parameters without changing the pre-trained model parameters. The above fine-tuning method not only maintains the high performance of the model, but also significantly improves the efficiency of model training and deployment.

[0107] Optionally, this embodiment uses the above LORA fine-tuning technology to perform model training, and the initialization parameter is the base model W0. For LORA, only the sub-model ΔW needs to be fine-tuned. Assume that the pre-training matrix is ​​W0∈R d×k , the update of the pre-training matrix can be expressed by the following formula:

[0108] W0+ΔW=W0+BA,B∈R d×r ,A∈R d×k

[0109] Among them, A can be used to represent the matrix, and B can be used to represent the training parameters. During the training process, W0 remains unchanged. When training starts, A is initialized with a random Gaussian distribution and B is initialized with a zero matrix. In the prediction stage, it can be expressed by the following formula:

[0110] h=W0x+ΔWx=W0x+BAx

[0111] Here, x can be used to represent the query, that is, W0 and ΔW are multiplied by the query x and finally added.

[0112] As an optional implementation, step S304, using prompt information samples and reply information samples, trains an information processing model, including: associating information inference rules with prompt information samples, wherein the information inference rules are used to represent the logical relationship for reasoning about prompt information; using prompt information samples associated with information inference rules, and reply information samples, trains an information processing model.

[0113] In this embodiment, after obtaining prompt information samples in the e-commerce scenario, the obtained prompt information samples can be associated with information inference rules. After the prompt information samples are associated with information inference rules, the prompt information samples associated with information inference rules and the reply information samples can be used to train an information processing model. Among them, the information inference rules can be used to represent the logical relationship for reasoning about the prompt information, so that the information processing model can realize the reasoning of complex tasks.

[0114] Optionally, the data format required for supervised fine-tuning is (prompt information, reply information). In order to improve the accuracy of reply information generated by the conversation understanding large model, after combining the query information and the semantic recognition information to obtain the prompt information, the CoT technology can be used in the prompt. Since CoT refers to a series of logically related thinking steps that can form a complete thinking process, the reply information can include two parts: the intermediate reasoning result and the answer, so that the conversation understanding large model can better realize the reasoning of complex tasks.

[0115] Optionally, in an online link, this embodiment recalls multiple segments of historical conversation information associated with the real question information of the current user through RAG, and assembles them into prompts as inquiries for the conversation understanding large model.

[0116] As an optional implementation, the method also includes: obtaining different types of inquiry information samples in an e-commerce scenario, wherein the type is used to indicate the inquiry intention corresponding to the inquiry information sample; obtaining manual reply information samples corresponding to the inquiry information samples, wherein the quality index of the manual reply information samples is higher than the quality index threshold; obtaining reply information samples matching the prompt information samples, including: utilizing the prompt information samples associated with information inference rules to combine the inquiry information samples and the manual reply information samples to generate reply information samples corresponding to the inquiry information samples.

[0117] In this embodiment, different types of inquiry information samples in an e-commerce scenario can be obtained. After obtaining different types of inquiry information samples in an e-commerce scenario, manual reply information samples corresponding to the inquiry information samples can be obtained. After associating the prompt information sample with the information inference rule and obtaining the prompt information sample associated with the information inference rule, the prompt information sample associated with the information inference rule is used to generate a reply information sample corresponding to the inquiry information sample by combining the inquiry information sample and the manual reply information sample. Among them, the type can be used to indicate the inquiry intention corresponding to the inquiry information sample. The quality index of the manual reply information sample is higher than the quality index threshold. The quality index threshold can be an index threshold set according to actual conditions, and the quality index threshold can be adjusted in real time according to actual needs.

[0118] Optionally, in order to ensure the comprehensiveness of the training data, this embodiment randomly extracts a certain number of users' real question information and manual customer service session information associated with the real question information for different inquiry intentions, i.e., question intentions, and uses GPT4 to query the optimized prompt to generate answer information and intermediate thinking corresponding to the real question information.

[0119] As an optional implementation, a prompt information sample associated with an information reasoning rule is used to generate a reply information sample corresponding to the query information sample by combining the query information sample and the manual reply information sample, including: when the prompt information sample associated with the information reasoning rule indicates that the reply information sample to be generated includes the answer sample of the query information sample, the prompt information sample associated with the information reasoning rule is used to obtain the answer sample of the query information sample from the manual reply information sample, and generate the reply information sample including the answer sample; when the prompt information sample associated with the information reasoning rule indicates that the reply information sample to be generated includes the intermediate reasoning result sample of the query information sample and the answer sample, the prompt information sample associated with the information reasoning rule is used to obtain the intermediate reasoning result sample of the query information sample and the answer sample from the manual reply information sample, and generate the reply information sample including the intermediate reasoning result sample and the answer sample, wherein the intermediate reasoning result sample is used to infer the answer sample.

[0120] In this embodiment, after the prompt information sample is associated with the information inference rule to obtain the prompt information sample associated with the information inference rule, when the prompt information sample associated with the information inference rule indicates that the reply information sample to be generated includes the answer sample of the inquiry information sample, the prompt information sample associated with the information inference rule is used to obtain the answer sample of the inquiry information sample from the artificial reply information sample, and generate the reply information sample including the answer sample. When the prompt information sample associated with the information inference rule indicates that the reply information sample to be generated includes the intermediate inference result sample of the inquiry information sample and the answer sample, the prompt information sample associated with the information inference rule is used to obtain the intermediate inference result sample of the inquiry information sample and the answer sample of the inquiry information sample from the artificial reply information sample, and generate the reply information sample including the intermediate inference result sample and the answer sample. Among them, the intermediate inference result sample can be used to infer the answer sample.

[0121] Optionally, after associating the prompt information sample with the information inference rule and obtaining the prompt information sample associated with the information inference rule, when the prompt sample associated with the information inference rule indicates that the reply information sample to be generated includes the intermediate inference result sample and the answer sample of the inquiry information sample, that is, the output of the intermediate thinking sample is required, then the type of training data is (prompt information sample-required to output the intermediate thinking sample, reply information sample). Therefore, by using the prompt information sample associated with the information inference rule, the intermediate inference result sample of the inquiry information sample and the answer sample of the inquiry information sample are obtained from the manual reply information sample, and the reply information sample including the intermediate inference result sample and the answer sample is generated.

[0122] Optionally, in the process of generating reply information samples, if intermediate thinking samples are generated, a large delay will be caused. Taking into account the performance requirements of the online model, this embodiment simultaneously constructs a part of prompt information samples and corresponding answer information samples that do not require the generation of intermediate thinking samples. In the case where the prompt sample associated with the information reasoning rule indicates that the reply information sample to be generated includes the answer sample of the inquiry information sample, that is, it is not required to output the intermediate thinking sample, then the type of training data is (prompt information sample-does not require the output of the intermediate thinking sample, reply information sample). Therefore, using the prompt information sample associated with the information reasoning rule, the answer sample of the inquiry information sample is obtained from the manual reply information sample, and the reply information sample including the answer sample is generated.

[0123] As an optional implementation, the method further includes: generating a reply information sample that does not include the answer sample when the answer sample fails to match the inquiry information sample.

[0124] In this embodiment, in order to maintain the ability of the large conversation understanding model to refuse to answer, a certain proportion of no-answer replies needs to be mixed into the training data. That is, when the answer sample fails to match the inquiry information sample, a reply information sample that does not include the answer sample can be generated.

[0125] As an optional implementation, an information processing model is trained using prompt information samples associated with information reasoning rules and reply information samples, including: using prompt information samples associated with information reasoning rules, prompt information samples not associated with information reasoning rules, and reply information samples to train an information processing model.

[0126] In this embodiment, the information processing model can be trained using prompt information samples associated with information reasoning rules, prompt information samples not associated with information reasoning rules, and reply information samples. The prompt information samples associated with information reasoning rules can be called CoT version annotated data, and the prompt information samples not associated with information reasoning rules can be called non-CoT version annotated data.

[0127] Optionally, this embodiment proposes a hybrid CoT fine-tuning strategy. That is, when training the model, the CoT version annotated data and the non-CoT version annotated data are mixed for training. In actual prediction, the non-CoT version prompt is used for prediction to ensure that the model accuracy is basically the same as the CoT version prompt, reduce the number of model output annotations (tokens), and thus reduce the online reasoning time.

[0128] In an embodiment of the present application, a prompt information sample in an e-commerce scenario is obtained, wherein the prompt information sample is used to prompt the generation of a reply information sample that matches the inquiry information sample in the prompt information sample; an information processing model is trained using the prompt information sample and the reply information sample, wherein the information processing model is guided by the prompt information to generate reply information that matches the inquiry information, the prompt information is generated based on the inquiry information and the historical conversation information associated with the inquiry information, the inquiry information is used to inquire about reply information associated with a product in the e-commerce scenario, the historical conversation information includes at least one historical inquiry information and historical reply information that matches the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive replies to the product, the quality index of the historical reply information is higher than the quality index threshold, thereby achieving the purpose of making the generated reply information closer to the user's actual expected answer, thereby achieving the improvement of the processing effect of the interactive information and solving the technical problem of poor processing effect of the interactive information.

[0129] The present application also provides another method for processing interactive information. Figure 4is a flowchart of a method for processing interactive information according to an embodiment of the present application. Figure 4 As shown, the method may include the following steps:

[0130] Step S402: Identify inquiry information to be answered on the e-commerce platform.

[0131] In the technical solution provided in the above step S402 of the present application, the inquiry information to be answered can be identified on the e-commerce platform. The inquiry information can be used to inquire about the reply information associated with the products on the e-commerce platform, for example, the inquiry information can be used to inquire about the quality, attributes and performance of the goods. The reply information matches the inquiry information, and the reply information can be information that replies to the inquiry information, for example, it can be high-quality reply information corresponding to the inquiry information generated by the intelligent customer service, which can be called intelligent reply information or answer information.

[0132] For example, the inquiry information may be real question information asked by the user, such as "What is the material of the product?", "What is the size of the product?", "How long will it take for the product to be shipped after purchase?", "Who is the product suitable for?" and other real questions asked by the user.

[0133] It should be noted that the above-mentioned commodities in the e-commerce scenario and the inquiry information associated with the commodities are only examples, and do not specifically limit the application scenario and the types of products and inquiry information in the application scenario.

[0134] Step S404: Acquire historical conversation information associated with the inquiry information.

[0135] In the technical solution provided in step S404 of the present application, after identifying the inquiry information to be answered, historical conversation information associated with the inquiry information can be obtained. The historical conversation information may include at least one historical inquiry information on the e-commerce platform and historical answer information matching the historical inquiry information.

[0136] In this embodiment, the above historical inquiry information may be information associated with the product inquired by the current user or other users in the historical period. The historical reply information may be information for replying to the historical inquiry information, for example, the historical reply information may be information for replying to the historical inquiry information related to the product by the manual customer service in the e-commerce scenario. The historical dialogue information may be called historical session information, and the historical session information may be the manual customer service historical session information.

[0137] It should be noted that the above historical inquiry information and historical reply information are generated for at least one round of interactive reply for the product, that is, the historical inquiry information and the historical reply information can constitute one or more groups of dialogue information. The quality index of the historical reply information is higher than the quality index threshold. Among them, the quality index threshold can be an index threshold set according to actual conditions, and the quality index threshold can be adjusted according to actual needs.

[0138] In this embodiment, after obtaining the inquiry information to be answered, historical conversation information associated with the inquiry information to be answered can be obtained. The historical conversation information can include one or more groups of historical inquiry information and historical reply information. The historical reply information can be information in which manual customer service replies to the historical inquiry information, or it can be high-quality information in which intelligent customer service replies to the historical inquiry information. This is only for illustration, and no specific limitation is imposed on the method of generating the historical reply information.

[0139] Step S406: Generate prompt information based on the query information and the historical dialogue information.

[0140] In the technical solution provided in the above step S406 of the present application, after obtaining the inquiry information to be replied and the historical conversation information associated with the inquiry information, prompt information is generated based on the obtained inquiry information and historical conversation information.

[0141] In this embodiment, after obtaining the inquiry information to be replied and the historical dialogue information associated with the inquiry information, the obtained inquiry information and the historical dialogue information associated with the inquiry information can be combined to obtain prompt information. Optionally, this embodiment combines the inquiry information and the historical dialogue information associated with the inquiry information according to an information combination format that matches the information processing model to obtain prompt information, so that the prompt information can be input into the information processing model.

[0142] Step S408, using the prompt information, guiding the information processing model to generate reply information on the e-commerce platform.

[0143] In the technical solution provided in the above step S408 of the present application, after the prompt information is generated based on the inquiry information and the historical conversation information, the prompt information is input into the information processing model, and the prompt information is used to guide the information processing model to generate reply information on the e-commerce platform. Among them, the information processing model can be a model trained using the prompt information samples and reply information samples associated with the product, which can be called a conversation understanding model. The prompt information sample can at least include an inquiry information sample that matches the reply information sample, which can be called a prompt word sample. The inquiry information sample can at least include historical inquiry information, and the reply information sample can at least include historical reply information, and the reply information sample can be called an answer sample.

[0144] In this embodiment, after obtaining the inquiry information to be replied, the obtained inquiry information can be used to obtain high-quality historical dialogue information associated with the inquiry information. After obtaining the high-quality historical dialogue information associated with the inquiry information, the obtained inquiry information and the high-quality historical dialogue information can be used as inputs of the information processing model to guide the information processing model to generate reply information that matches the inquiry information. Optionally, the reply information is the reply information after optimizing and supplementing the intelligent answer information using the historical conversation information of the manual customer service. The reply information can be closer to the user's actual expected answer, thereby improving the processing effect of the interactive information.

[0145] Through the above steps S402 to S408 of the present application, on the e-commerce platform, inquiry information to be answered is identified, wherein the inquiry information is used to inquire about reply information associated with the product on the e-commerce platform; historical conversation information associated with the inquiry information is obtained, wherein the historical conversation information includes at least one historical inquiry information on the e-commerce platform, and historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive replies to the product, and the quality index of the historical reply information is higher than the quality index threshold; prompt information is generated based on the inquiry information and the historical conversation information; using the prompt information, the information processing model is guided to generate reply information on the e-commerce platform, wherein the reply information matches the inquiry information, wherein the information processing model is trained using prompt information samples and reply information samples generated on the e-commerce platform, and the prompt information samples include inquiry information samples matching the reply information samples, thereby achieving the purpose of making the generated reply information closer to the user's real expected answer, thereby achieving the improvement of the processing effect of the interactive information, and solving the technical problem of poor processing effect of the interactive information.

[0146] The present application also provides another method for processing interactive information. Figure 5 is a flowchart of a method for processing interactive information according to an embodiment of the present application. Figure 5 As shown, the method may include the following steps:

[0147] Step S502, responding to the inquiry instruction on the operation interface, and displaying the inquiry information to be answered on the operation interface.

[0148] In the technical solution provided in the above step S502 of the present application, in response to the inquiry instruction acting on the operation interface, the inquiry information to be answered can be displayed on the operation interface. Among them, the operation interface can be a user operation interface, and the type of the operation interface is not specifically limited here. The inquiry instruction can be an instruction to inquire about relevant information on the operation interface. The inquiry information can be used to inquire about the reply information associated with the product on the e-commerce platform. For example, the inquiry information can be used to inquire about the quality, attributes and performance of the goods. The reply information matches the inquiry information, and the reply information can be information that replies to the inquiry information. For example, it can be high-quality reply information corresponding to the inquiry information generated by intelligent customer service, which can be called intelligent reply information or answer information.

[0149] For example, the inquiry information may be real question information asked by the user, such as "What is the material of the product?", "What is the size of the product?", "How long will it take for the product to be shipped after purchase?", "Who is the product suitable for?" and other real questions asked by the user.

[0150] It should be noted that the above-mentioned commodities in the e-commerce scenario and the inquiry information associated with the commodities are only examples, and do not specifically limit the application scenario and the types of products and inquiry information in the application scenario.

[0151] Step S504, in response to the information reply instruction acting on the operation interface, display the historical dialogue information associated with the inquiry information on the operation interface.

[0152] In the technical solution provided in the above step S504 of the present application, after the inquiry information to be answered is displayed on the operation interface, in response to the information answering instruction acting on the operation interface, the historical dialogue information associated with the inquiry information is displayed on the operation interface. The information answering instruction may be an instruction to answer the inquiry information displayed on the operation interface. The historical dialogue information may include at least one historical inquiry information on the e-commerce platform, and historical answer information matching the historical inquiry information.

[0153] In this embodiment, the above historical inquiry information may be information associated with the product inquired by the current user or other users in the historical period. The historical reply information may be information for replying to the historical inquiry information, for example, the historical reply information may be information for replying to the historical inquiry information related to the product by the manual customer service in the e-commerce scenario. The historical dialogue information may be called historical session information, and the historical session information may be the manual customer service historical session information.

[0154] It should be noted that the above historical inquiry information and historical reply information are generated for at least one round of interactive reply for the product, that is, the historical inquiry information and the historical reply information can constitute one or more groups of dialogue information. The quality index of the historical reply information is higher than the quality index threshold. Among them, the quality index threshold can be an index threshold set according to actual conditions, and the quality index threshold can be adjusted according to actual needs.

[0155] In this embodiment, after the inquiry information to be answered is displayed on the operation interface, in response to the information reply instruction on the operation interface, historical conversation information associated with the inquiry information is displayed on the operation interface. The historical conversation information may include one or more groups of historical inquiry information and historical reply information. The historical reply information may be information provided by manual customer service in response to the historical inquiry information, or it may be high-quality information provided by intelligent customer service in response to the historical inquiry information. This is only for illustration, and no specific limitation is imposed on the method of generating the historical reply information.

[0156] Step S506: display prompt information on the operation interface.

[0157] In the technical solution provided in the above step S506 of the present application, after the historical dialogue information associated with the inquiry information is displayed on the operation interface in response to the information reply instruction acting on the operation interface, the prompt information can be displayed on the operation interface. The prompt information can be generated based on the inquiry information and the historical dialogue information.

[0158] In this embodiment, after the inquiry information to be answered and the historical dialogue information associated with the inquiry information are displayed on the operation interface, the displayed inquiry information and the historical dialogue information associated with the inquiry information can be combined to obtain prompt information. Optionally, this embodiment combines the inquiry information and the historical dialogue information associated with the inquiry information according to an information combination format that matches the information processing model to obtain prompt information, so that the prompt information can be input into the information processing model.

[0159] Step S508: displaying reply information matching the inquiry information on the operation interface.

[0160] In the technical solution provided in the above step S508 of the present application, after the prompt information is displayed on the operation interface, the reply information matching the inquiry information can be displayed on the operation interface. The reply information can be generated by guiding the information processing model using the prompt information. The information processing model can be a model trained using the prompt information samples and reply information samples associated with the product, which can be called a conversation understanding model. The prompt information samples can at least include inquiry information samples matching the reply information samples, which can be called prompt word samples. The inquiry information samples can at least include historical inquiry information, and the reply information samples can at least include historical reply information, and the reply information samples can be called answer samples.

[0161] In this embodiment, after the inquiry information to be answered is displayed on the operation interface, high-quality historical dialogue information associated with the inquiry information can be displayed on the operation interface. After the high-quality historical dialogue information associated with the inquiry information is displayed, the displayed inquiry information and the high-quality historical dialogue information can be used as inputs of the information processing model to guide the information processing model to generate reply information that matches the inquiry information. Optionally, the reply information is the reply information that optimizes and supplements the intelligent answer information using the historical conversation information of the manual customer service, and the reply information can be closer to the user's actual expected answer, thereby improving the processing effect of the interactive information.

[0162] Through the above steps S502 to S508 of the present application, in response to the inquiry instruction on the operation interface, the inquiry information to be answered is displayed on the operation interface, wherein the inquiry information is used to inquire about the reply information associated with the product; in response to the information reply instruction on the operation interface, the historical dialogue information associated with the inquiry information is displayed on the operation interface, wherein the historical dialogue information includes at least one historical inquiry information and historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive reply for the product, and the quality index of the historical reply information is higher than the quality index threshold; prompt information is displayed on the operation interface, wherein the prompt information is generated based on the inquiry information and the historical dialogue information; reply information matching the inquiry information is displayed on the operation interface, wherein the reply information is generated by guiding the information processing model using the prompt information, the information processing model is trained using the prompt information samples and reply information samples associated with the product, and the prompt information samples include inquiry information samples matching the reply information samples, thereby achieving the purpose of making the generated reply information closer to the user's real expected answer, thereby achieving the improvement of the processing effect of the interactive information and solving the technical problem of poor processing effect of the interactive information.

[0163] The present application also provides another method for processing interactive information. Figure 6is a flowchart of a method for processing interactive information according to an embodiment of the present application. Figure 6 As shown, the method may include the following steps:

[0164] Step S602: Acquire inquiry information to be answered by calling the first interface.

[0165] In the technical solution provided in the above step S602 of the present application, the inquiry information to be answered can be obtained by calling the first interface. The first interface may include a first parameter, and the parameter value of the first parameter may include inquiry information, and the inquiry information may be used to inquire about the reply information associated with the product on the e-commerce platform, for example, the inquiry information may be used to inquire about the quality, attributes and performance of the product. The reply information matches the inquiry information, and the reply information may be information that replies to the inquiry information, for example, it may be high-quality reply information corresponding to the inquiry information generated by the intelligent customer service, which may be called intelligent reply information or answer information.

[0166] For example, the inquiry information may be real question information asked by the user, such as "What is the material of the product?", "What is the size of the product?", "How long will it take for the product to be shipped after purchase?", "Who is the product suitable for?" and other real questions asked by the user.

[0167] It should be noted that the above-mentioned commodities in the e-commerce scenario and the inquiry information associated with the commodities are only examples, and do not specifically limit the application scenario and the types of products and inquiry information in the application scenario.

[0168] Step S604: Acquire historical conversation information associated with the inquiry information.

[0169] In the technical solution provided in the above step S604 of the present application, after obtaining the inquiry information to be answered by calling the first interface, historical conversation information associated with the inquiry information can be obtained. The historical conversation information may include at least one historical inquiry information on the e-commerce platform, and historical answer information matching the historical inquiry information.

[0170] In this embodiment, the above historical inquiry information may be information associated with the product inquired by the current user or other users in the historical period. The historical reply information may be information for replying to the historical inquiry information, for example, the historical reply information may be information for replying to the historical inquiry information related to the product by the manual customer service in the e-commerce scenario. The historical dialogue information may be called historical session information, and the historical session information may be the manual customer service historical session information.

[0171] It should be noted that the above historical inquiry information and historical reply information are generated for at least one round of interactive reply for the product, that is, the historical inquiry information and the historical reply information can constitute one or more groups of dialogue information. The quality index of the historical reply information is higher than the quality index threshold. Among them, the quality index threshold can be an index threshold set according to actual conditions, and the quality index threshold can be adjusted according to actual needs.

[0172] In this embodiment, after obtaining the inquiry information to be answered by calling the first interface, historical conversation information associated with the inquiry information can be obtained. The historical conversation information can include one or more groups of historical inquiry information and historical reply information. The historical reply information can be information in which manual customer service replies to the historical inquiry information, or it can be high-quality information in which intelligent customer service replies to the historical inquiry information. This is only for illustration, and no specific limitation is imposed on the method of generating the historical reply information.

[0173] Step S606: Generate prompt information based on the query information and the historical dialogue information.

[0174] In the technical solution provided in the above step S606 of the present application, after acquiring the historical conversation information associated with the inquiry information, prompt information can be generated based on the inquiry information and the historical conversation information. The prompt information can be generated based on the inquiry information and the historical conversation information.

[0175] In this embodiment, after obtaining the inquiry information to be answered and the historical dialogue information associated with the inquiry information, the obtained inquiry information and the historical dialogue information associated with the inquiry information can be combined to obtain prompt information. Optionally, this embodiment combines the inquiry information and the historical dialogue information associated with the inquiry information according to an information combination format that matches the information processing model to obtain prompt information, so that the prompt information can be input into the information processing model.

[0176] Step S608: using the prompt information, guiding the information processing model to generate reply information matching the inquiry information.

[0177] In the technical solution provided in the above step S608 of the present application, after the prompt information is generated based on the inquiry information and the historical conversation information, the prompt information is input into the information processing model, and the prompt information is used to guide the information processing model to generate reply information matching the inquiry information. Among them, the reply information can be generated by guiding the information processing model using the prompt information. The information processing model can be a model trained using the prompt information samples and reply information samples associated with the product, which can be called a conversation understanding model. The prompt information samples can at least include inquiry information samples matching the reply information samples, which can be called prompt word samples. The inquiry information samples can at least include historical inquiry information, and the reply information samples can at least include historical reply information, and the reply information samples can be called answer samples.

[0178] In this embodiment, after obtaining the inquiry information to be answered, high-quality historical dialogue information associated with the inquiry information can be obtained. After obtaining the high-quality historical dialogue information associated with the inquiry information, the inquiry information and the high-quality historical dialogue information can be used as inputs of the information processing model to guide the information processing model to generate reply information that matches the inquiry information. Optionally, the reply information is the reply information that optimizes and supplements the intelligent answer information using the historical conversation information of the manual customer service. The reply information can be closer to the user's actual expected answer, thereby improving the processing effect of the interactive information.

[0179] Step S610: output reply information by calling the second interface.

[0180] In the technical solution provided in the above step S610 of the present application, after using the prompt information to guide the information processing model to generate reply information matching the inquiry information, the reply information can be output by calling the second interface. The second interface can include a second parameter, and the parameter value of the second parameter can include the reply information.

[0181] Through the above steps S602 to S610 of the present application, the inquiry information to be answered is obtained by calling the first interface, wherein the first interface includes a first parameter, and the parameter value of the first parameter includes inquiry information, and the inquiry information is used to inquire about the reply information associated with the product; historical dialogue information associated with the inquiry information is obtained, wherein the historical dialogue information includes at least one historical inquiry information and historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive reply to the product, and the quality index of the historical reply information is higher than the quality index threshold; based on the inquiry information and the historical dialogue information, prompt information is generated; using the prompt information, the information processing model is guided to generate reply information matching the inquiry information, wherein the information processing model is trained using prompt information samples and reply information samples associated with the product, and the prompt information samples include inquiry information samples matching the reply information samples; the reply information is output by calling the second interface, wherein the second interface includes a second parameter, and the parameter value of the second parameter includes reply information, thereby achieving the purpose of making the generated reply information closer to the user's real expected answer, thereby achieving the improvement of the processing effect of the interactive information, and solving the technical problem of poor processing effect of the interactive information.

[0182] 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 application, for example, the data for verification, 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.

[0183] At present, in e-commerce scenarios, it is necessary to give accurate responses to users' real questions based on multi-source information such as product attributes, product details pages, and reviews. Intelligent customer service has been widely used in the e-commerce field due to its advantages such as high efficiency, low cost, and 24-hour response. However, this type of knowledge usually has a certain amount of noise, low quality, and can only solve a limited number of problems. On the other hand, the manual side can give comprehensive answers based on various types of knowledge such as multimodality and web searches, and the answering language is relatively friendly, so the satisfaction of manual customer service is higher than that of intelligent customer service.

[0184] Furthermore, in e-commerce scenarios, the answer information extracted based on product attributes, detail pages, and other information is different from the distribution of the user's real question information, resulting in the inability of intelligent customer service to extract relevant answer information. Secondly, the language style and organization of information such as product attributes and detail pages tend to be objective and concise, resulting in a significant gap between the answer information extracted from them and the response information expected by users. Based on the user's real question information, combined with product knowledge and world knowledge, manual customer service has generated a large amount of high-quality answer information in historical conversations with consumers.

[0185] In the related art, the knowledge of the historical conversation information is extracted by performing a Cartesian product on the real question information of the consumer or user and the answer information of the customer service in a historical conversation information, and then the product knowledge is filtered out offline through the QA semantic matching model and put into the product knowledge base. The extracted answer information is returned online through similarity calculation and other retrieval methods based on the real question information of the user. However, this method has no contextual understanding ability for historical conversation information, and the accuracy of QA knowledge extracted by QA text matching is low. At the same time, the generated answer information is only the single-sentence answer information that has appeared in the historical conversation information. There is no ability to summarize the answer information of multiple rounds of customer service based on historical conversations, resulting in poor integrity and quality of the generated answer information. In addition, the offline link knowledge production method cannot utilize historical conversation information in real time, and relies on the QA knowledge retrieval ability, resulting in low effectiveness and utilization of the generated answer information, resulting in poor processing effect of interactive information.

[0186] However, in the embodiment of the present application, in order to make good use of the large amount of conversation data replied by these manual customer service, an efficient online feedback solution for manual customer service conversation data based on a large language model is proposed. The relevant historical conversation information is recalled according to the user's real problem information, and high-quality answer information is generated based on the large model, and replied to consumers in real time, truly achieving a closed loop between manual customer service and intelligence.

[0187] That is to say, in order to make full use of the historical conversation information of manual customer service, this embodiment proposes an efficient online feedback solution for manual customer service conversation data based on a large language model, which realizes the real-time generation of accurate answer information through historical conversation data, and significantly improves the output rate and quality of the answer information generated by intelligent customer service, thereby improving the user experience.

[0188] The above method of this embodiment is further introduced below.

[0189] Figure 7 is a schematic diagram of guiding a large conversation understanding model to generate intelligent answers according to an embodiment of the present application, such as Figure 7As shown, the user question asked by the current user is asked to the conversation understanding big model, and the historical conversation record of the product related to the user question is obtained at the same time. The historical conversation record consists of the user question asked by the current user or other users at the historical moment and the manual answer given by the manual customer service to the user question. That is, the user question asked by the current user and the historical conversation record associated with the user question are used as the query of the conversation understanding big model, so as to realize the real-time optimization and supplement of the intelligent answer by the manual customer service conversation data through the conversation understanding big model, which greatly reduces the cost of manual customer service, and the answers generated based on the real manual customer service conversation are closer to the real expected answers of consumers, which significantly improves the pre-sales satisfaction.

[0190] Figure 8 is a schematic diagram of recalling historical session records according to an embodiment of the present application, such as Figure 8 As shown, the user questions asked by the current user are vectorized. In addition, multiple historical conversation data (shown in the figure are the first historical conversation data, the second historical conversation data, and the third historical conversation data) are screened, cleaned, and deduplicated through intent recognition, invalid speech recognition, clustering algorithms, etc. Further, an index is constructed for each historical conversation data, that is, multiple product consultation questions related to each historical conversation data are extracted from each historical conversation data as the index of the historical conversation data. Figure 8 As shown in the figure, the indexes of the first historical conversation data are q1 and q2, the indexes of the second historical conversation data are q3 and q4, and the index of the third historical conversation data is q5. The open source StructBERT FAQ question-answering model is used to vectorize the user's real questions (queries) and historical conversation queries in the knowledge base and then recall them. In the online link, multiple historical conversation records related to the real questions asked by the current user are recalled through RAG, assembled into prompts, and used as queries for the large conversation understanding model to generate intelligent answers.

[0191] Based on the user's question information and the historical conversation information of the manual customer service related to the question information, the large model can generate answer information corresponding to the question information. First, the base model needs to have a strong ability to understand e-commerce scene dialogues and be able to summarize accurate and high-quality answer information from the dialogue. Secondly, the base model needs to have a low hallucination rate and should refuse to answer questions that cannot be accurately answered based on the content of the dialogue. Finally, the base model needs a certain ability to follow instructions and can give answer information in a specified format.

[0192] In summary, we can determine that the base model is qwen1.5 7B, and select LORA as the fine-tuning method. LORA can be used to add low-order matrices and learn new task-specific parameters without changing the pre-trained model parameters. The above fine-tuning method not only maintains the high performance of the model, but also significantly improves the efficiency of model training and deployment.

[0193] The data format required by SFT is (prompt information, reply information). In order to improve the accuracy of reply information generated by the conversation understanding model, CoT technology is added to prompt, so that the conversation understanding model can better realize the reasoning of complex tasks. Therefore, the reply information can contain two parts: the intermediate reasoning result and the answer.

[0194] In order to ensure the comprehensiveness of the training data, this embodiment randomly extracts a certain number of users' real question information and manual customer service conversation information associated with the real question information for different inquiry intentions, i.e., question intentions, and uses GPT4 to query the optimized prompt to generate answer information and intermediate thinking corresponding to the real question information. In order to maintain the ability of the large conversation understanding model to refuse to answer, a certain proportion of unanswered replies needs to be mixed into the training data. At the same time, in the process of generating reply information samples, if intermediate thinking samples are generated, it will cause a large delay. Taking into account the performance requirements of the online model, this embodiment also constructs a part of prompt information samples and corresponding answer information samples that do not require the generation of intermediate thinking samples.

[0195] In summary, the training data of this embodiment mainly includes the following two types: (prompt information - no requirement to output intermediate thinking, reply information), (prompt information - requirement to output intermediate thinking, reply information).

[0196] This embodiment uses the above LORA fine-tuning technology to train the model. The initialization parameter is the base model W0. For LORA, only the sub-model ΔW needs to be fine-tuned. Assume that the pre-training matrix is ​​W0∈R d×k , the update of the pre-training matrix can be expressed by the following formula:

[0197] W0+ΔW=W0+BA,B∈R d×r ,A∈R d×k

[0198] Among them, A can be used to represent the matrix, and B can be used to represent the training parameters. During the training process, W0 remains unchanged. When training starts, A is initialized with a random Gaussian distribution and B is initialized with a zero matrix. In the prediction stage, it can be expressed by the following formula:

[0199] h=W0x+ΔWx=W0x+BAx

[0200] Here, x can be used to represent the query, that is, W0 and ΔW are multiplied by the query x and finally added.

[0201] Furthermore, this embodiment proposes a hybrid CoT fine-tuning strategy. That is, when training the model, the CoT version annotated data and the non-CoT version annotated data are mixed for training. In actual prediction, the non-CoT version prompt is used for prediction to ensure that the model accuracy is basically the same as the CoT version prompt, reduce the number of model output tokens, and thus reduce the online reasoning time.

[0202] This embodiment implements a conversation understanding module that recalls historical customer service conversations based on the RAG link and generates large model answers in real time, which can make more efficient use of customer service conversation data. At the same time, a hybrid CoT fine-tuning solution is proposed, and the conversation understanding large model is trained in combination with the LORA fine-tuning technology to dynamically recall the artificial conversation data according to user questions, and then generate high-quality answers. Compared with the e-commerce scenario that does not use the conversation understanding large model, the hit rate of the reply information generated by this embodiment is increased by 4 percentage points (percentage point, abbreviated as pt), the market satisfaction is increased by 3pt, and the inquiry conversion rate is increased by 0.3pt.

[0203] This embodiment constructs intelligent customer service conversation understanding training data, proposes a hybrid CoT fine-tuning solution, and combines the LORA fine-tuning technology to train the conversation understanding large model, thereby improving the large model's conversation understanding and answer generation capabilities. It also uses RAG to recall historical conversations of manual customer service in real time to generate better answers, thereby achieving a data closed loop between manual customer service data and intelligent customer service, improving user satisfaction and bringing about an increase in inquiry conversion rate.

[0204] In an embodiment of the present application, query information to be answered is obtained, wherein the query information is used to inquire about answer information associated with a product; historical conversation information associated with the query information is obtained, wherein the historical conversation information includes at least one historical query information and historical answer information matching the historical query information, the historical query information and the historical answer information are generated for at least one round of interactive answering for the product, and the quality index of the historical answer information is higher than the quality index threshold; prompt information is generated based on the query information and the historical conversation information; using the prompt information, an information processing model is guided to generate answer information matching the query information, wherein the information processing model is trained using prompt information samples and answer information samples associated with the product, and the prompt information samples include query information samples matching the answer information samples. That is to say, the embodiment of the present application obtains high-quality historical conversation information associated with the query information to be answered, and uses the high-quality historical conversation information to optimize and supplement the answer information matching the query information, thereby achieving the purpose of making the generated answer information closer to the user's real expected answer, thereby achieving the improvement of the processing effect of the interactive information, and solving the technical problem of poor processing effect of the interactive information.

[0205] Fig. 9 is a structural block diagram of a computing environment of a method for processing interactive information according to an embodiment of the present application, such as Fig. 9 As shown, computing environment 901 includes multiple computing nodes (such as servers) running on a distributed network (shown as 910-1, 910-2, ...) in the figure. The computing nodes all contain local processing and memory resources, and end users 902 can remotely run applications or store data in computing environment 901. Applications can be provided as multiple services 920-1, 920-2, 920-3 and 920-4 in computing environment 901, representing services "A", "D", "E" and "H" respectively.

[0206] The end user 902 can provide and access services through a web browser or other software application on the client, and in some embodiments, the end user 902's provision and / or request can be provided to the entry gateway 930. The entry gateway 930 may include a corresponding agent to handle the provision and / or request for services (one or more services provided in the computing environment 901).

[0207] Services are provided or deployed based on various virtualization technologies supported by the computing environment 901. In some embodiments, services can be provided based on virtualization based on virtual machines (VMs for short), container-based virtualization, and / or similar methods. Virtual machine-based virtualization can be to simulate a real computer by initializing a virtual machine, and execute 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 for short) so that multiple workloads can run on a single operating system instance.

[0208] In an embodiment based on container virtualization, several containers of a service can be assembled into a Pod (e.g., a Kubernetes Pod). Fig. 9 As shown, service 920-2 can be equipped with one or more Pods 940-1, 940-2, ..., 940-N (collectively referred to as Pods). Pods can include a proxy 945 and one or more containers 942-1, 942-2, ..., 942-M (collectively referred to as containers). One or more containers in a Pod process requests related to one or more corresponding functions of the service, and the proxy 945 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.

[0209] During operation, executing a user request from the end user 902 may require calling one or more services in the computing environment 901, and executing one or more functions of a service may require calling one or more functions of another service. Fig. 9 As shown, service "A" 920-1 receives a user request from an end user 902 from an ingress gateway 930, service "A" 920-1 may call service "D" 920-2, and service "D" 920-2 may request service "E" 920-3 to perform one or more functions.

[0210] The computing environment described above can be a cloud computing environment, where the allocation of resources is managed by the cloud service provider, allowing the development of functions without considering the implementation, adjustment or expansion of servers. The computing environment allows developers to execute code in response to events without building or maintaining complex infrastructure. Services can be divided into a set of functions that can be automatically and independently scaled, rather than expanding a single hardware device to handle potential loads.

[0211] According to an embodiment of the present application, there is also provided a method for implementing the above Figure 2 The interactive information processing method and the interactive information processing device shown are provided.

[0212] Fig.10 is a schematic diagram of a device for processing interactive information according to an embodiment of the present application, such as Fig.10 As shown, the interactive information processing device 1000 may include: a first acquiring unit 1002 , a second acquiring unit 1004 , a first generating unit 1006 and a second generating unit 1008 .

[0213] The first acquisition unit 1002 is used to acquire inquiry information to be answered, wherein the inquiry information is used to inquire about answer information associated with the product.

[0214] The second acquisition unit 1004 is used to acquire historical conversation information associated with the inquiry information, wherein the historical conversation information includes at least one historical inquiry information and historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive reply for the product, and the quality index of the historical reply information is higher than the quality index threshold.

[0215] The first generating unit 1006 is used to generate prompt information based on the query information and the historical dialogue information.

[0216] The second generation unit 1008 is used to use the prompt information to guide the information processing model to generate reply information matching the inquiry information, wherein the information processing model is trained using prompt information samples and reply information samples associated with the product, and the prompt information samples include inquiry information samples matching the reply information samples.

[0217] It should be noted that the first acquisition unit 1002, the second acquisition unit 1004, the first generation unit 1006 and the second generation unit 1008 correspond to steps S202 to S208, and the four units and the corresponding steps implement the same examples and application scenarios, but are not limited to the above disclosed content. It should be noted that the above units can be hardware components or software components stored in a memory (e.g., memory 1504) and processed by one or more processors (e.g., processors 1502a, 1502b..., 1502n), and the above units can also be run in a computer terminal A as part of the device.

[0218] In the interactive information processing device, the inquiry information to be answered is obtained by the first acquisition unit 1002, wherein the inquiry information is used to inquire about the answer information associated with the product. The historical dialogue information associated with the inquiry information is obtained by the second acquisition unit 1004, wherein the historical dialogue information includes at least one historical inquiry information and historical answer information matching the historical inquiry information, the historical inquiry information and the historical answer information are generated for at least one round of interactive answering for the product, and the quality index of the historical answer information is higher than the quality index threshold. The prompt information is generated by the first generation unit 1006 based on the inquiry information and the historical dialogue information. The prompt information is used by the second generation unit 1008 to guide the information processing model to generate the answer information matching the inquiry information, wherein the information processing model is trained by using the prompt information samples and the answer information samples associated with the product, and the prompt information samples include the inquiry information samples matching the answer information samples, so as to achieve the purpose of making the generated answer information closer to the user's real expected answer, thereby achieving the improvement of the processing effect of the interactive information and solving the technical problem of poor processing effect of the interactive information.

[0219] According to an embodiment of the present application, there is also provided a method for implementing the above Figure 3 The model determination method shown is a model determination device.

[0220] Fig.11 is a schematic diagram of a model determination device according to an embodiment of the present application, such as Fig.11 As shown, the model determination device 1100 may include: a third acquisition unit 1102 and a training unit 1104.

[0221] The third acquisition unit 1102 is used to acquire a prompt information sample in an e-commerce scenario, wherein the prompt information sample is used to prompt the generation of a reply information sample that matches the inquiry information sample in the prompt information sample.

[0222] The training unit 1104 is used to train an information processing model using prompt information samples and reply information samples, wherein the information processing model is guided by the prompt information to generate reply information matching the inquiry information, the prompt information is generated based on the inquiry information and historical conversation information associated with the inquiry information, the inquiry information is used to inquire about reply information associated with a product in an e-commerce scenario, the historical conversation information includes at least one historical inquiry information and historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive replies to the product, and the quality index of the historical reply information is higher than the quality index threshold.

[0223] Here, the third acquisition unit 1102 and the training unit 908 correspond to steps S302 to S304, and the two units and the corresponding steps implement the same examples and application scenarios, but are not limited to the above disclosed content. It should be noted that the above unit can be a hardware component or software component stored in a memory (e.g., memory 1504) and processed by one or more processors (e.g., processors 1502a, 1502b..., 1502n), and the above unit can also be run in a computer terminal A as part of the device.

[0224] In the determination device of the model, a prompt information sample in an e-commerce scenario is obtained by a third acquisition unit 1102, wherein the prompt information sample is used to prompt the generation of a reply information sample matching the inquiry information sample in the prompt information sample. The information processing model is trained by a training unit 1104 using the prompt information sample and the reply information sample, wherein the information processing model is guided by the prompt information to generate reply information matching the inquiry information, the prompt information is generated based on the inquiry information and the historical dialogue information associated with the inquiry information, the inquiry information is used to inquire about the reply information associated with the product in the e-commerce scenario, the historical dialogue information includes at least one historical inquiry information and the historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive replies for the product, and the quality index of the historical reply information is higher than the quality index threshold, thereby achieving the purpose of making the generated reply information closer to the user's real expected answer, thereby achieving the improvement of the processing effect of the interactive information and solving the technical problem of poor processing effect of the interactive information.

[0225] According to an embodiment of the present application, there is also provided a method for implementing the above Figure 4 The interactive information processing method and the interactive information processing device shown are provided.

[0226] Fig.12 is a schematic diagram of another interactive information processing device according to an embodiment of the present application, such as Fig.12 As shown, the interactive information processing device 1200 may include: an identification unit 1202 , a fourth acquisition unit 1204 , a third generation unit 1206 and a fourth generation unit 1208 .

[0227] The identification unit 1202 is used to identify inquiry information to be answered on the e-commerce platform, wherein the inquiry information is used to inquire about answer information associated with the product on the e-commerce platform.

[0228] The fourth acquisition unit 1204 is used to obtain historical conversation information associated with the inquiry information, wherein the historical conversation information includes at least one historical inquiry information on the e-commerce platform, and historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive replies for the product, and the quality index of the historical reply information is higher than the quality index threshold.

[0229] The third generating unit 1206 is used to generate prompt information based on the query information and the historical dialogue information.

[0230] The fourth generation unit 1208 is used to use the prompt information to guide the information processing model to generate reply information on the e-commerce platform, wherein the reply information matches the inquiry information, wherein the information processing model is trained using the prompt information samples and reply information samples generated on the e-commerce platform, and the prompt information samples include inquiry information samples that match the reply information samples.

[0231] It should be noted that the identification unit 1202, the fourth acquisition unit 1204, the third generation unit 1206 and the fourth generation unit 1208 correspond to steps S402 to S408, and the four units and the corresponding steps implement the same examples and application scenarios, but are not limited to the above disclosed content. It should be noted that the above units can be hardware components or software components stored in a memory (e.g., memory 1504) and processed by one or more processors (e.g., processors 1502a, 1502b..., 1502n), and the above units can also be run in a computer terminal A as part of the device.

[0232] In the interactive information processing device, the identification unit 1202 identifies the inquiry information to be answered on the e-commerce platform, wherein the inquiry information is used to inquire about the answer information associated with the product on the e-commerce platform. The fourth acquisition unit 1204 acquires the historical dialogue information associated with the inquiry information, wherein the historical dialogue information includes at least one historical inquiry information on the e-commerce platform, and the historical answer information matching the historical inquiry information, the historical inquiry information and the historical answer information are generated for at least one round of interactive answering for the product, and the quality index of the historical answer information is higher than the quality index threshold. The third generation unit 1206 generates prompt information based on the inquiry information and the historical dialogue information. The fourth generation unit 1208 uses the prompt information to guide the information processing model to generate answer information on the e-commerce platform, wherein the answer information matches the inquiry information, wherein the information processing model is trained using the prompt information samples and the answer information samples generated on the e-commerce platform, and the prompt information samples include the inquiry information samples matching the answer information samples, thereby achieving the purpose of making the generated answer information closer to the user's real expected answer, thereby achieving the improvement of the processing effect of the interactive information, and solving the technical problem of poor processing effect of the interactive information.

[0233] Fig.13 is a schematic diagram of another interactive information processing device according to an embodiment of the present application, such as Fig.13 As shown, the interactive information processing device 1300 may include: a first display unit 1302 , a second display unit 1304 , a third display unit 1306 and a fourth display unit 1308 .

[0234] The first display unit 1302 is used to respond to the inquiry instruction on the operation interface and display the inquiry information to be answered on the operation interface, wherein the inquiry information is used to inquire about the answer information associated with the product.

[0235] The second display unit 1304 is used to respond to the information reply instruction acting on the operation interface, and display the historical dialogue information associated with the inquiry information on the operation interface, wherein the historical dialogue information includes at least one historical inquiry information, and historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive reply for the product, and the quality index of the historical reply information is higher than the quality index threshold.

[0236] The third display unit 1306 is used to display prompt information on the operation interface, wherein the prompt information is generated based on the query information and the historical dialogue information.

[0237] The fourth display unit 1308 is used to display reply information matching the inquiry information on the operation interface, wherein the reply information is generated by using prompt information to guide the information processing model, the information processing model is trained using prompt information samples and reply information samples associated with the product, and the prompt information samples include inquiry information samples matching the reply information samples.

[0238] It should be noted that the first display unit 1302, the second display unit 1304, the third display unit 1306 and the fourth display unit 1308 correspond to steps S502 to S508, and the four units and the corresponding steps implement the same examples and application scenarios, but are not limited to the above disclosed content. It should be noted that the above units can be hardware components or software components stored in a memory (e.g., memory 1504) and processed by one or more processors (e.g., processors 1502a, 1502b..., 1502n), and the above units can also be run in computer terminal A as part of the device.

[0239] In the interactive information processing device, the first display unit 1302 responds to the inquiry instruction on the operation interface, and displays the inquiry information to be answered on the operation interface, wherein the inquiry information is used to inquire about the answer information associated with the product. The second display unit 1304 responds to the information answer instruction on the operation interface, and displays the historical dialogue information associated with the inquiry information on the operation interface, wherein the historical dialogue information includes at least one historical inquiry information and the historical answer information matching the historical inquiry information, the historical inquiry information and the historical answer information are generated for at least one round of interactive answering for the product, and the quality index of the historical answer information is higher than the quality index threshold. The third display unit 1306 displays the prompt information on the operation interface, wherein the prompt information is generated based on the inquiry information and the historical dialogue information. The fourth display unit 1308 displays reply information matching the inquiry information on the operation interface, wherein the reply information is generated by using the prompt information to guide the information processing model, the information processing model is trained using the prompt information samples and reply information samples associated with the product, and the prompt information samples include inquiry information samples matching the reply information samples, thereby achieving the purpose of making the generated reply information closer to the user's actual expected answer, thereby improving the processing effect of the interactive information and solving the technical problem of poor processing effect of the interactive information.

[0240] Fig.14 is a schematic diagram of another interactive information processing device according to an embodiment of the present application, such as Fig.14 As shown, the interactive information processing device 1400 may include: a fifth acquiring unit 1402 , a sixth acquiring unit 1404 , a fifth generating unit 1406 , a sixth generating unit 1408 and an output unit 1410 .

[0241] The fifth acquisition unit 1402 is used to acquire the inquiry information to be answered by calling the first interface, wherein the first interface includes a first parameter, the parameter value of the first parameter includes inquiry information, and the inquiry information is used to inquire about the answer information associated with the product.

[0242] The sixth acquisition unit 1404 is used to acquire historical conversation information associated with the inquiry information, wherein the historical conversation information includes at least one historical inquiry information and historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive replies for the product, and the quality index of the historical reply information is higher than the quality index threshold.

[0243] The fifth generating unit 1406 is used to generate prompt information based on the query information and the historical dialogue information.

[0244] The sixth generation unit 1408 is used to use the prompt information to guide the information processing model to generate reply information matching the inquiry information, wherein the information processing model is trained using prompt information samples and reply information samples associated with the product, and the prompt information samples include inquiry information samples matching the reply information samples.

[0245] The output unit 1410 is used to output the reply information by calling the second interface, wherein the second interface includes a second parameter, and the parameter value of the second parameter includes the reply information.

[0246] It should be noted that the fifth acquisition unit 1402, the sixth acquisition unit 1404, the fifth generation unit 1406, the sixth generation unit 1408 and the output unit 1410 correspond to steps S602 to S610, and the five units and the corresponding steps implement the same examples and application scenarios, but are not limited to the above disclosed content. It should be noted that the above units can be hardware components or software components stored in a memory (e.g., memory 1504) and processed by one or more processors (e.g., processors 1502a, 1502b..., 1502n), and the above units can also be run in a computer terminal A as part of the device.

[0247] In the interactive information processing device, the query information to be answered is obtained by calling the first interface through the fifth acquisition unit 1402, wherein the first interface includes a first parameter, and the parameter value of the first parameter includes the query information, and the query information is used to inquire about the reply information associated with the product. The historical dialogue information associated with the query information is obtained by the sixth acquisition unit 1404, wherein the historical dialogue information includes at least one historical query information and the historical reply information matching the historical query information, the historical query information and the historical reply information are generated for at least one round of interactive reply for the product, and the quality index of the historical reply information is higher than the quality index threshold. The prompt information is generated by the fifth generation unit 1406 based on the query information and the historical dialogue information. The prompt information is used by the sixth generation unit 1408 to guide the information processing model to generate the reply information matching the query information, wherein the information processing model is trained by using the prompt information sample and the reply information sample associated with the product, and the prompt information sample includes the query information sample matching the reply information sample. The reply information is outputted by calling the second interface through the output unit 1410, wherein the second interface includes a second parameter, and the parameter value of the second parameter includes the reply information, thereby achieving the purpose of making the generated reply information closer to the user's actual expected answer, thereby improving the processing effect of the interactive information and solving the technical problem of poor processing effect of the interactive information.

[0248] The embodiment of the present application 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.

[0249] Optionally, in this embodiment, the computer terminal may be located in at least one network device among a plurality of network devices of the computer network.

[0250] In this embodiment, the above-mentioned computer terminal can execute the program code of the following steps in the method for processing interactive information: obtaining inquiry information to be answered, wherein the inquiry information is used to inquire about reply information associated with the product; obtaining historical conversation information associated with the inquiry information, wherein the historical conversation information includes at least one historical inquiry information, and historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive replies to the product, and the quality index of the historical reply information is higher than the quality index threshold; generating prompt information based on the inquiry information and the historical conversation information; using the prompt information, guiding the information processing model to generate reply information matching the inquiry information, wherein the information processing model is trained using prompt information samples and reply information samples associated with the product, and the prompt information samples include inquiry information samples matching the reply information samples.

[0251] Optionally, Fig.15 is a structural block diagram of a computer terminal according to an embodiment of the present application, such as Fig.15 As shown, the computer terminal A may include: one or more (only one is shown in the figure) processors 1502 , a memory 1504 and a transmission device 1506 .

[0252] Among them, the memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the interactive information processing method and device in the embodiment of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, realizing the above-mentioned interactive information processing method. 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 arranged relative to the processor, and these remote memories can be connected to the 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.

[0253] Optionally, the processor may also execute the program code of the following steps: searching the external information repository of the information processing model according to the query information to obtain historical conversation information, wherein the external information repository includes conversation information in addition to prompt information samples and reply information samples, and the conversation information includes different types of knowledge about the product.

[0254] Optionally, the processor may also execute the program code of the following steps: vectorizing the query information; searching an external information repository according to the vectorized query information to obtain search results, and determining historical conversation information corresponding to the search results, wherein the search results are used to represent the vectorized historical conversation information.

[0255] Optionally, the processor may also execute program code for the following steps: searching an external information repository according to the vectorized query information to obtain vectorized historical query information associated with the product; searching an external information repository according to the vectorized historical query information to obtain retrieval results including the vectorized historical query information and the vectorized historical reply information.

[0256] Optionally, the processor may also execute the program code of the following steps: using the vectorized historical query information as index information, indexing in an external information repository a retrieval result including the vectorized historical query information and the vectorized historical reply information.

[0257] Optionally, the processor may also execute program codes of the following steps: performing semantic recognition on the context in the historical conversation information to obtain semantic recognition information; and combining the inquiry information and the semantic recognition information to obtain prompt information.

[0258] Optionally, the processor may also execute the following program code: controlling the information processing model, reasoning the prompt information according to the information reasoning rules, and obtaining the reply information, wherein the information reasoning rules are used to represent the logical relationship of the information processing model in reasoning the prompt information.

[0259] Optionally, the processor may also execute program code for the following steps: when the prompt information indicates that the reply information to be generated includes an answer to the inquiry information, the information processing model is controlled to infer the prompt information according to information reasoning rules, and the reply information including the answer to the inquiry information is output; when the prompt information indicates that the reply information to be generated includes an intermediate reasoning result and an answer to the inquiry information, the information processing model is controlled to infer the prompt information according to information reasoning rules, and the reply information including the intermediate reasoning result and the answer is output, wherein the intermediate reasoning result is used to enable the information processing model to infer the answer.

[0260] Optionally, the processor may also execute program code of the following steps: in the case where the answer fails to match the query information, the information processing model may be controlled to output reply information that does not include the answer.

[0261] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: obtain prompt information samples in the e-commerce scenario, wherein the prompt information samples are used to prompt the generation of reply information samples that match the inquiry information samples in the prompt information samples; use the prompt information samples and the reply information samples to train an information processing model, wherein the information processing model is guided by the prompt information to generate reply information that matches the inquiry information, the prompt information is generated based on the inquiry information and the historical dialogue information associated with the inquiry information, the inquiry information is used to inquire about the reply information associated with the product in the e-commerce scenario, the historical dialogue information includes at least one historical inquiry information, and the historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive replies to the product, and the quality index of the historical reply information is higher than the quality index threshold.

[0262] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: on the e-commerce platform, identify the inquiry information to be answered, wherein the inquiry information is used to inquire about the reply information associated with the product on the e-commerce platform; obtain historical conversation information associated with the inquiry information, wherein the historical conversation information includes at least one historical inquiry information on the e-commerce platform, and historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive replies to the product, and the quality index of the historical reply information is higher than the quality index threshold; based on the inquiry information and the historical conversation information, generate prompt information; use the prompt information to guide the information processing model to generate reply information on the e-commerce platform, wherein the reply information matches the inquiry information, wherein the information processing model is trained using prompt information samples and reply information samples generated on the e-commerce platform, and the prompt information samples include inquiry information samples matching the reply information samples.

[0263] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: responding to the inquiry instruction acting on the operation interface, displaying the inquiry information to be answered on the operation interface, wherein the inquiry information is used to inquire about the reply information associated with the product; responding to the information reply instruction acting on the operation interface, displaying the historical dialogue information associated with the inquiry information on the operation interface, wherein the historical dialogue information includes at least one historical inquiry information and the historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive replies for the product, and the quality index of the historical reply information is higher than the quality index threshold; displaying the prompt information on the operation interface, wherein the prompt information is generated based on the inquiry information and the historical conversation information; displaying the reply information matching the inquiry information on the operation interface, wherein the reply information is generated by using the prompt information to guide the information processing model, the information processing model is trained using the prompt information samples and the reply information samples associated with the product, and the prompt information samples include the inquiry information samples matching the reply information samples.

[0264] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: obtain the inquiry information to be answered by calling the first interface, wherein the first interface includes a first parameter, and the parameter value of the first parameter includes inquiry information, and the inquiry information is used to inquire about the reply information associated with the product; obtain historical conversation information associated with the inquiry information, wherein the historical conversation information includes at least one historical inquiry information, and historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive replies to the product, and the quality index of the historical reply information is higher than the quality index threshold; generate prompt information based on the inquiry information and the historical conversation information; use the prompt information to guide the information processing model to generate reply information matching the inquiry information, wherein the information processing model is trained using prompt information samples and reply information samples associated with the product, and the prompt information samples include inquiry information samples matching the reply information samples; output the reply information by calling the second interface, wherein the second interface includes a second parameter, and the parameter value of the second parameter includes the reply information.

[0265] It can be understood by those skilled in the art that Fig.15 The structure shown is for illustration only, and the computer terminal A may also be a smart phone (such as an Android phone, an iOS phone, etc.), a tablet computer, a PDA, a mobile Internet device (Mobile Internet Devices, abbreviated as MID), a personal access display (Personal Access Display, abbreviated as PAD), and other terminal devices. Fig.15 It does not limit the structure of the above-mentioned computer terminal A. For example, the computer terminal A may also include Fig.15 More or fewer components (such as network interfaces, display devices, etc.) shown in, or having Fig.15 Different configurations are shown.

[0266] A person of ordinary skill in the art can 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, and the storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, etc.

[0267] The embodiment of the present application 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 processing interaction information provided in the embodiment.

[0268] Optionally, in this embodiment, the computer-readable storage medium may be located in any one of the computer terminals in a computer terminal group in a computer network, or in any one of the mobile terminals in a mobile terminal group.

[0269] Optionally, in this embodiment, the computer-readable storage medium is configured to store program code for performing the following steps: obtaining inquiry information to be answered, wherein the inquiry information is used to inquire about reply information associated with the product; obtaining historical conversation information associated with the inquiry information, wherein the historical conversation information includes at least one historical inquiry information and historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive replies to the product, and a quality indicator of the historical reply information is higher than a quality indicator threshold; generating prompt information based on the inquiry information and the historical conversation information; using the prompt information, guiding an information processing model to generate reply information matching the inquiry information, wherein the information processing model is trained using prompt information samples and reply information samples associated with the product, and the prompt information samples include inquiry information samples matching the reply information samples.

[0270] Optionally, the computer-readable storage medium may also execute program code for the following steps: searching an external information repository of an information processing model according to query information to obtain historical conversation information, wherein the external information repository includes conversation information in addition to prompt information samples and reply information samples, and the conversation information includes different types of knowledge about the product.

[0271] Optionally, the computer-readable storage medium may also execute program code for the following steps: vectorizing the query information; searching an external information repository according to the vectorized query information to obtain search results, and determining historical conversation information corresponding to the search results, wherein the search results are used to represent the vectorized historical conversation information.

[0272] Optionally, the computer-readable storage medium may also execute program code for the following steps: searching an external information repository according to the vectorized query information to obtain vectorized historical query information associated with the product; searching an external information repository according to the vectorized historical query information to obtain retrieval results including the vectorized historical query information and the vectorized historical reply information.

[0273] Optionally, the computer-readable storage medium may also execute program code for the following steps: using the vectorized historical query information as index information, indexing in an external information repository a retrieval result including the vectorized historical query information and the vectorized historical reply information.

[0274] Optionally, the computer-readable storage medium may also execute program code for the following steps: performing semantic recognition on the context in the historical conversation information to obtain semantic recognition information; and combining the query information and the semantic recognition information to obtain prompt information.

[0275] Optionally, the computer-readable storage medium may also execute program code of the following steps: controlling the information processing model, reasoning the prompt information according to information reasoning rules, and obtaining reply information, wherein the information reasoning rules are used to represent the logical relationship of the information processing model in reasoning the prompt information.

[0276] Optionally, the computer-readable storage medium may also execute program code for the following steps: when the prompt information indicates that the reply information to be generated includes an answer to the inquiry information, controlling the information processing model to infer the prompt information according to information reasoning rules, and outputting reply information including the answer to the inquiry information; when the prompt information indicates that the reply information to be generated includes an intermediate reasoning result and an answer to the inquiry information, controlling the information processing model to infer the prompt information according to information reasoning rules, and outputting reply information including the intermediate reasoning result and the answer, wherein the intermediate reasoning result is used to enable the information processing model to infer the answer.

[0277] Optionally, the computer-readable storage medium may also execute program code for the following steps: in the event that the answer fails to match the query information, the information processing model may be controlled to output reply information that does not include the answer.

[0278] As an optional example, a computer-readable storage medium is configured to store program code for performing the following steps: obtaining prompt information samples in an e-commerce scenario, wherein the prompt information samples are used to prompt the generation of reply information samples that match the inquiry information samples in the prompt information samples; using the prompt information samples and the reply information samples to train an information processing model, wherein the information processing model is guided by the prompt information to generate reply information that matches the inquiry information, the prompt information is generated based on the inquiry information and historical conversation information associated with the inquiry information, the inquiry information is used to inquire about reply information associated with a product in the e-commerce scenario, the historical conversation information includes at least one historical inquiry information, and historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive replies to the product, and the quality index of the historical reply information is higher than the quality index threshold.

[0279] As an optional example, a computer-readable storage medium is configured to store program code for performing the following steps: on an e-commerce platform, identifying inquiry information to be answered, wherein the inquiry information is used to inquire about reply information associated with a product on the e-commerce platform; obtaining historical conversation information associated with the inquiry information, wherein the historical conversation information includes at least one historical inquiry information on the e-commerce platform, and historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information being generated for at least one round of interactive replies to the product, and a quality indicator of the historical reply information being higher than a quality indicator threshold; generating prompt information based on the inquiry information and the historical conversation information; and using the prompt information to guide an information processing model to generate reply information on the e-commerce platform, wherein the reply information matches the inquiry information, wherein the information processing model is trained using prompt information samples and reply information samples generated on the e-commerce platform, and the prompt information samples include inquiry information samples matching the reply information samples.

[0280] As an optional example, a computer-readable storage medium is configured to store program code for performing the following steps: in response to an inquiry instruction acting on an operation interface, displaying inquiry information to be answered on the operation interface, wherein the inquiry information is used to inquire about reply information associated with a product; in response to an information reply instruction acting on the operation interface, displaying historical conversation information associated with the inquiry information on the operation interface, wherein the historical conversation information includes at least one historical inquiry information and historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive replies to the product, and a quality indicator of the historical reply information is higher than a quality indicator threshold; displaying prompt information on the operation interface, wherein the prompt information is generated based on the inquiry information and the historical conversation information; displaying reply information matching the inquiry information on the operation interface, wherein the reply information is generated by using the prompt information to guide an information processing model, the information processing model is trained using prompt information samples and reply information samples associated with the product, and the prompt information samples include inquiry information samples matching the reply information samples.

[0281] As an optional example, a computer-readable storage medium is configured to store program code for performing the following steps: obtaining inquiry information to be answered by calling a first interface, wherein the first interface includes a first parameter, and a parameter value of the first parameter includes inquiry information, and the inquiry information is used to inquire about reply information associated with a product; obtaining historical conversation information associated with the inquiry information, wherein the historical conversation information includes at least one historical inquiry information, and historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive replies to the product, and a quality indicator of the historical reply information is higher than a quality indicator threshold; generating prompt information based on the inquiry information and the historical conversation information; using the prompt information, guiding an information processing model to generate reply information matching the inquiry information, wherein the information processing model is trained using prompt information samples and reply information samples associated with the product, and the prompt information samples include inquiry information samples matching the reply information samples; outputting reply information by calling a second interface, wherein the second interface includes a second parameter, and a parameter value of the second parameter includes reply information.

[0282] The embodiment of the present application further provides a computer program product. Optionally, in this embodiment, the computer program product may include a computer program, and the computer program implements the method provided in the embodiment when executed by a processor.

[0283] Optionally, the computer program product may include a non-volatile computer-readable storage medium, and the non-volatile computer-readable storage medium may be used to store a computer program. When the computer program is executed by a processor, the method provided in the above embodiment is implemented.

[0284] An embodiment of the present application may provide an electronic device, which may include a memory and a processor.

[0285] Fig.16 It is a block diagram of an electronic device according to a method for processing interactive information according to an embodiment of the present application. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, 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 application described and / or required herein.

[0286] like Fig.16As shown, the device 1600 includes a computing unit 1601, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1602 or a computer program loaded from a storage unit 1608 into a random access memory (RAM) 1603. In the RAM 1603, various programs and data required for the operation of the device 1600 can also be stored. The computing unit 1601, the ROM 1602, and the RAM 1603 are connected to each other via a bus 1604. An input / output (I / O) interface 1605 is also connected to the bus 1604.

[0287] A number of components in the device 1600 are connected to the I / O interface 1605, including: an interrogation unit 1606, such as a keyboard, a mouse, etc.; an output unit 1607, such as various types of displays, speakers, etc.; a storage unit 1608, such as a disk, an optical disk, etc.; and a communication unit 1609, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 1609 allows the device 1600 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0288] The computing unit 1601 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 1601 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 units running machine learning model algorithms, digital signal processors (DSP), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 1601 performs the various methods and processes described above, such as a processing method for interactive information. For example, in some embodiments, the processing method for interactive information may be implemented as a computer software program, which is tangibly included in a machine-readable medium, such as a storage unit 1608. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 1600 via the ROM 1602 and / or the communication unit 1609. When the computer program is loaded into RAM 1603 and executed by computing unit 1601, one or more steps of the method for processing interaction information described above may be performed. Alternatively, in other embodiments, computing unit 1601 may be configured to execute the method for processing interaction information in any other appropriate manner (e.g., by means of firmware).

[0289] Various implementations of the systems and techniques described above herein may be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard parts (ASSPs), system-on-a-chip systems (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations may 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 that may be a special purpose or general purpose programmable processor that may receive data and instructions from a storage system, at least one interrogation device, and at least one output device, and transmit data and instructions to the storage system, the at least one interrogation device, and the at least one output device.

[0290] The program code for implementing the method of the present application can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code can be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0291] The method embodiment provided in Embodiment 1 of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Fig.17 is a hardware structure block diagram of a computer terminal (or mobile device) for implementing a method for processing interactive information according to an embodiment of the present application, such as Fig.17As shown, the computer terminal 170 (or mobile device) may include one or more (shown in the figure as 1702a, 1702b, ..., 1702n) processors 1702 (the processor 1702 may include but is not limited to a processing device such as a microprocessor (Microcontroller Unit, abbreviated as MCU) or a programmable logic device (Field Programmable Gate Array, abbreviated as FPGA)), a memory 1704 for storing data, and a transmission device 1706 for communication functions. In addition, it may also include: a display, an inquiry / output interface (I / O interface), a universal serial bus (Universal Ser ia l Bus, abbreviated as 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 can be understood by those skilled in the art that Fig.17 The structure shown is only for illustration and does not limit the structure of the above electronic device. Fig.17 More or fewer components as shown, or with Fig.17 Different configurations are shown.

[0292] Fig.17 The hardware structure block diagram shown can be used not only as an exemplary block diagram of the above-mentioned computer terminal 170 (or mobile device), but also as an exemplary block diagram of the above-mentioned server. In an optional embodiment, Fig.17 The block diagram shows the use of the above Fig.17 The computer terminal 170 (or mobile device) is shown as an embodiment of a computing node in the computing environment 701.

[0293] In the context of the present application, a machine-readable medium may be a tangible medium that may contain or store a program for use by an instruction execution system, device or equipment or for use in combination with an instruction execution system, device or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may 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 may 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 for short), an optical fiber, a portable compact disk read-only memory (CD-ROM for short), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0294] To provide interaction with a user, the systems and techniques described herein may be implemented on a computer having: a display device (e.g., a cathode ray tube (CRT) or a liquid crystal display (LCD), a monitor for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a path ball), through which the user can provide queries to the computer. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and queries from the user may be received in any form (including acoustic queries, voice queries, or tactile queries).

[0295] 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 with 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.

[0296] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0297] It should be noted that the serial numbers of the above-mentioned embodiments of the present application are only for description and do not represent the advantages or disadvantages of the embodiments.

[0298] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0299] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0300] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0301] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0302] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, 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, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, read-only memories, random access memories, mobile hard disks, magnetic disks or optical disks.

[0303] The above are only preferred implementations of the present application. 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 application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for processing interactive information, characterized in that: include: Acquire inquiry information to be answered, wherein the inquiry information is used to inquire about answer information associated with the product; Acquire historical dialogue information associated with the inquiry information, wherein the historical dialogue information includes at least one historical inquiry information and historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive reply for the product, and a quality indicator of the historical reply information is higher than a quality indicator threshold; Generate prompt information based on the query information and the historical conversation information; Utilizing the prompt information, guide the information processing model to generate the reply information matching the inquiry information, wherein the information processing model is trained using the prompt information samples and reply information samples associated with the product, and the prompt information samples include inquiry information samples matching the reply information samples.

2. The method according to claim 1, characterized in that: Acquiring historical conversation information associated with the query information, including: According to the query information, a search is performed in the external information repository of the information processing model to obtain the historical dialogue information, wherein the external information repository includes dialogue information other than the prompt information sample and the reply information sample, and the dialogue information includes different types of knowledge about the product.

3. The method according to claim 2, characterized in that According to the query information, searching is performed in the external information library of the information processing model to obtain the historical conversation information, including: Performing vectorization processing on the query information; According to the vectorized query information, a search is performed in the external information database to obtain a search result, and the historical conversation information corresponding to the search result is determined, wherein the search result is used to represent the vectorized historical conversation information.

4. The method according to claim 3, characterized in that According to the vectorized query information, a search is performed in the external information database to obtain a search result, including: According to the vectorized query information, searching is performed in the external information database to obtain the vectorized historical query information associated with the product; According to the vectorized historical query information, a search is performed in the external information database to obtain the search result including the vectorized historical query information and the vectorized historical reply information.

5. The method according to claim 4, characterized in that According to the vectorized historical query information, a search is performed in the external information database to obtain the search result including the vectorized historical query information and the vectorized historical reply information, including: The vectorized historical query information is used as index information to index the retrieval result including the vectorized historical query information and the vectorized historical reply information in the external information repository.

6. The method according to any one of claims 1 to 5, characterized in that Generate prompt information based on the query information and the historical conversation information, including: Performing semantic recognition on the context in the historical conversation information to obtain semantic recognition information; The query information and the semantic recognition information are combined to obtain the prompt information.

7. The method according to claim 6, characterized in that Using the prompt information, guiding the information processing model to generate reply information matching the inquiry information, including: The information processing model is controlled to infer the prompt information according to information inference rules to obtain the reply information, wherein the information inference rules are used to represent the logical relationship of the information processing model in inferring the prompt information.

8. The method according to claim 7, characterized in that Controlling the information processing model to infer the prompt information according to information inference rules to obtain the reply information includes: When the prompt information indicates that the reply information to be generated includes an answer to the inquiry information, controlling the information processing model to infer the prompt information according to the information inference rule and outputting the reply information including the answer to the inquiry information; When the prompt information indicates that the reply information to be generated includes the intermediate reasoning result of the inquiry information and the answer, the information processing model is controlled to reason the prompt information according to the information reasoning rule, and the reply information including the intermediate reasoning result and the answer is output, wherein the intermediate reasoning result is used to enable the information processing model to infer the answer.

9. The method according to claim 8, characterized in that The method further comprises: In a case where the answer fails to match the query information, the information processing model is controlled to output the reply information without including the answer.

10. A method for determining a model, characterized in that: include: Acquire a prompt information sample in an e-commerce scenario, wherein the prompt information sample is used to prompt the generation of a reply information sample that matches the inquiry information sample in the prompt information sample; An information processing model is trained using the prompt information samples and the reply information samples, wherein the information processing model is guided by the prompt information to generate reply information matching the inquiry information, the prompt information is generated based on the inquiry information and historical conversation information associated with the inquiry information, the inquiry information is used to inquire about the reply information associated with the product in the e-commerce scenario, the historical conversation information includes at least one historical inquiry information and historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive replies for the product, and the quality indicator of the historical reply information is higher than the quality indicator threshold.

11. The method according to claim 10, characterized in that Using the prompt information sample and the reply information sample, training an information processing model includes: Determining a pre-trained model in the e-commerce scenario, wherein the pre-trained model is used to represent a large language model that allows training in the e-commerce scenario; Adding a sub-model to the pre-trained model, wherein the sub-model is used to represent a matrix whose order is less than an order threshold; Using the prompt information sample and the answer information sample, training the sub-model; The pre-trained model and the trained sub-model are determined as the information processing model.

12. The method according to claim 10, characterized in that Using the prompt information sample and the reply information sample, training an information processing model includes: Associating the prompt information sample with an information inference rule, wherein the information inference rule is used to represent a logical relationship for inferring the prompt information; The information processing model is trained by using the prompt information samples associated with the information inference rules and the reply information samples.

13. The method according to claim 12, characterized in that The method further comprises: Acquire different types of query information samples in the e-commerce scenario, wherein the types are used to indicate query intentions corresponding to the query information samples; Acquire a manual reply information sample corresponding to the inquiry information sample, wherein a quality index of the manual reply information sample is higher than the quality index threshold; Acquiring a reply information sample matching the prompt information sample includes: using the prompt information sample associated with the information inference rule, combining the query information sample and the manual reply information sample to generate the reply information sample corresponding to the query information sample.

14. The method according to claim 13, characterized in that Using the prompt information sample associated with the information inference rule, the query information sample and the manual reply information sample are combined to generate the reply information sample corresponding to the query information sample, including: In a case where the prompt information sample associated with the information inference rule indicates that the reply information sample to be generated includes an answer sample of the inquiry information sample, using the prompt information sample associated with the information inference rule, obtaining an answer sample of the inquiry information sample from the manual reply information sample, and generating the reply information sample including the answer sample; In the case where the prompt information sample associated with the information inference rule indicates that the reply information sample to be generated includes the intermediate inference result sample of the inquiry information sample and the answer sample, the prompt information sample associated with the information inference rule is used to obtain the intermediate inference result sample of the inquiry information sample and the answer sample of the inquiry information sample from the manual reply information sample, and generate the reply information sample including the intermediate inference result sample and the answer sample, wherein the intermediate inference result sample is used to infer the answer sample.

15. The method according to claim 14, characterized in that The method further comprises: In the case where the answer sample fails to match the inquiry information sample, the reply information sample is generated without including the answer sample.

16. The method according to claim 14, characterized in that The information processing model is trained by using the prompt information sample associated with the information inference rule and the reply information sample, including: The information processing model is trained by using the prompt information samples associated with the information inference rule, the prompt information samples not associated with the information inference rule, and the reply information samples.

17. A method for processing interactive information, characterized in that: include: On an e-commerce platform, identifying inquiry information to be answered, wherein the inquiry information is used to ask for answer information associated with a product on the e-commerce platform; Acquire historical dialogue information associated with the inquiry information, wherein the historical dialogue information includes at least one historical inquiry information on the e-commerce platform and historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated by at least one round of interactive reply for the product, and a quality indicator of the historical reply information is higher than a quality indicator threshold; Generate prompt information based on the query information and the historical conversation information; The prompt information is used to guide an information processing model to generate the reply information on the e-commerce platform, wherein the reply information matches the inquiry information. The information processing model is trained using the prompt information samples and reply information samples generated on the e-commerce platform, and the prompt information samples include inquiry information samples that match the reply information samples.

18. A method for processing interactive information, characterized in that: include: In response to an inquiry instruction acting on an operation interface, inquiry information to be answered is displayed on the operation interface, wherein the inquiry information is used to inquire about answer information associated with the product; In response to an information reply instruction acting on the operation interface, displaying historical dialogue information associated with the inquiry information on the operation interface, wherein the historical dialogue information includes at least one historical inquiry information and historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive reply for the product, and the quality index of the historical reply information is higher than the quality index threshold; Displaying prompt information on the operation interface, wherein the prompt information is generated based on the query information and the historical dialogue information; The reply information matching the inquiry information is displayed on the operation interface, wherein the reply information is generated by using the prompt information to guide an information processing model, the information processing model is trained using prompt information samples and reply information samples associated with the product, and the prompt information samples include inquiry information samples matching the reply information samples.

19. A method for processing interactive information, characterized in that: include: Acquire inquiry information to be answered by calling a first interface, wherein the first interface includes a first parameter, a parameter value of the first parameter includes the inquiry information, and the inquiry information is used to inquire about answer information associated with the product; Acquire historical dialogue information associated with the inquiry information, wherein the historical dialogue information includes at least one historical inquiry information and historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive reply for the product, and a quality indicator of the historical reply information is higher than a quality indicator threshold; Generate prompt information based on the query information and the historical conversation information; Using the prompt information, guiding an information processing model to generate the reply information matching the inquiry information, wherein the information processing model is trained using prompt information samples and reply information samples associated with the product, and the prompt information samples include inquiry information samples matching the reply information samples; The reply information is output by calling a second interface, wherein the second interface includes a second parameter, and a parameter value of the second parameter includes the reply information.

20. A system for processing interactive information, characterized in that: include: A client, used to upload inquiry information to be answered, wherein the inquiry information is used to inquire about answer information associated with the product; The server is used to obtain historical conversation information associated with the inquiry information, wherein the historical conversation information includes at least one historical inquiry information and historical reply information matching the historical inquiry information, the historical inquiry information and the historical reply information are generated for at least one round of interactive reply for the product, and the quality index of the historical reply information is higher than the quality index threshold; based on the inquiry information and the historical conversation information, generate prompt information; using the prompt information, guide an information processing model to generate the reply information matching the inquiry information, wherein the information processing model is trained using prompt information samples and reply information samples associated with the product, and the prompt information samples include inquiry information samples matching the reply information samples; and send the reply information to the client.

21. An electronic device, characterized in that: include: A memory storing an executable program; A processor, configured to run the program, wherein the program executes the method according to any one of claims 1 to 19 when running.

22. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored executable program, wherein when the executable program is executed, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 19.

23. A computer program product, characterized in that The invention comprises a computer program which, when executed by a processor, implements the method according to any one of claims 1 to 19.

Images