Response generation method and reply generation device based on large-scale language model

Abstract

The invention provides a reply generation method and device based on a large-scale language model, electronic equipment and a non-instantaneous computer readable storage medium, and the reply generation method comprises the steps: responding to a question instruction, and determining whether a question included in the question instruction is an uncertain question; under the condition that the problem is determined to be the uncertainty problem, generating a task corresponding to the problem; and according to the generated task, performing an online session based on the large-scale language model to collect information corresponding to the task so as to generate a reply corresponding to the question. According to the embodiment of the invention, the online session is carried out in the form of converting the question of the user into the task, so that the information corresponding to the task is collected, the reply corresponding to the question is generated, the reply accuracy of the intelligent customer service is improved, and the pressure of the manual customer service is relieved.

Description

Technical Field

[0001] The present application relates to the field of intelligent customer service, and more specifically, to a reply generation method and reply generation device, an electronic device, and a non-transitory computer-readable storage medium based on a large-scale language model. Background Art

[0002] The entity roles of wholesale websites include buyers, platform stores and suppliers. Due to their relatively special application scenarios, there are also relatively complex inquiry relationships on the inquiry side, such as Figure 1 shown.

[0003] At present, the customer service of large stores mainly connects with buyers and upstream suppliers. When there is no intelligent customer service, there is only manual customer service. The main problems are: the cost of manual customer service is too high, the reception quality is not high, and the reception efficiency is low. With the emergence of intelligent customer service, although the pressure of manual customer service has been alleviated, the judgment of conditions such as orders and users when determining the response plan is very complex and changeable. Therefore, in actual operation, the traditional intelligent customer service response method cannot provide actual solutions or replies. Summary of the invention

[0004] The present application aims to propose a reply generation method and reply generation device, electronic device, and non-transitory computer-readable storage medium based on a large-scale language model to solve the problem of low accuracy of the solutions provided by current intelligent customer service.

[0005] According to one aspect of the present application, a method for generating an answer based on a large-scale language model is proposed, comprising: in response to a question instruction, determining whether the question included in the question instruction is an uncertain question; in the case where it is determined that the question is an uncertain question, generating a task corresponding to the question; based on the generated task, conducting an online conversation based on the large-scale language model to collect information corresponding to the task, thereby generating an answer corresponding to the question.

[0006] According to some embodiments, the question instruction includes a question input by a user, and determining whether the question included in the question instruction is an uncertain question includes: performing intent recognition based on the input question to determine whether the question included in the question instruction is an uncertain question.

[0007] According to some embodiments, generating a task corresponding to the problem includes: generating the task corresponding to the problem using a preset data format.

[0008] According to some embodiments, according to the generated task, based on the large-scale language model, an online conversation is conducted to collect information corresponding to the task, so as to generate a reply corresponding to the question, including: according to the task, using a preset rule engine to determine the rule corresponding to the task; based on the large-scale language model, generating an information collection task according to the rule; according to the generated information collection task, conducting at least one online conversation to collect information corresponding to the task, so as to generate a reply corresponding to the question.

[0009] According to some embodiments, generating a task corresponding to the question includes: using an online conversation to collect question requirements or corresponding answer content; using a preset knowledge base to generate an answer corresponding to the question; and / or filling in content in a preset reply template based on the input question to generate an answer corresponding to the input question.

[0010] According to some embodiments, the task generated corresponding to the question includes a plurality of tasks, and the answer generation method further includes: generating an answer corresponding to the input question based on answer data collected by the plurality of tasks.

[0011] According to some embodiments, the answer generation method also includes: conducting an online conversation with the user to determine whether the user problem is solved; in response to the user's continue execution instruction, continuing to generate a reply corresponding to the new question instruction based on the new question instruction input by the user until the user's end of the conversation instruction is received.

[0012] According to one aspect of the present application, a reply generation device based on a large-scale language model is proposed, comprising: a question feature determination unit, for determining, in response to a question instruction, whether the question included in the question instruction is an uncertain question; a task generation unit, for generating a task corresponding to the question when it is determined that the question is an uncertain question; and a reply generation unit, for conducting an online conversation based on the generated task and the large-scale language model to collect information corresponding to the task, thereby generating a reply corresponding to the question.

[0013] According to one aspect of the present application, an electronic device is proposed, comprising: a processor; a memory for storing a computer program; when the computer program is executed by the processor, the processor implements the method described in any of the previous embodiments.

[0014] According to one aspect of the present application, a non-transitory computer-readable storage medium is provided, on which computer-readable instructions are stored, and when the instructions are executed by a processor, the processor executes the method described in any of the above embodiments.

[0015] According to an embodiment of the present application, an online conversation is conducted by converting user questions into tasks to collect information corresponding to the tasks and generate replies corresponding to the questions, thereby improving the accuracy of intelligent customer service responses and alleviating the pressure on manual customer service.

[0016] It should be understood that the foregoing general description and the following detailed description are exemplary only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for describing the embodiments. By describing the exemplary embodiments in detail with reference to the drawings, the above and other objectives, features and advantages of the present application will become more apparent.

[0018] Figure 1 A schematic diagram of a customer service process for a website after approval is shown.

[0019] Figure 2 A flow chart of a method for generating a reply based on a large-scale language model according to an exemplary embodiment of the present application is shown.

[0020] Figure 3 A schematic diagram of an intelligent customer service workflow according to an example embodiment of the present application is shown.

[0021] Figure 4 A schematic diagram of a workflow of a rule engine according to an exemplary embodiment of the present application is shown.

[0022] Figure 5 A schematic diagram of an intent recognition result according to an exemplary embodiment of the present application is shown.

[0023] Figure 6 A schematic diagram of order status information according to an exemplary embodiment of the present application is shown.

[0024] Figure 7 A schematic diagram of content filling according to an exemplary embodiment of the present application is shown.

[0025] Figure 8 A block diagram of a reply generation device based on a large-scale language model according to an exemplary embodiment of the present application is shown.

[0026] Fig. 9 An electronic device according to an exemplary embodiment of the present application is shown. DETAILED DESCRIPTION

[0027] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that this application will be comprehensive and complete and will fully convey the concepts of the example embodiments to those skilled in the art. In the figures, the same figures show the same or similar parts, and thus their repeated description will be omitted.

[0028] 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 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.

[0029] The described features, structures or characteristics may be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure may be practiced without one or more of these specific details, or other modes, components, materials, devices or operations may be adopted. In these cases, known structures, methods, devices, implementations, materials or operations will not be shown or described in detail.

[0030] The flowcharts shown in the accompanying drawings are only exemplary and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps can be decomposed, and some operations / steps can be combined or partially combined, so the actual execution order may change according to actual conditions.

[0031] The terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices.

[0032] Before describing the embodiments of the present application, the terms appearing in the present application are first explained.

[0033] Big store: refers to merchants with high sales volume and popularity on e-commerce platforms.

[0034] Large-scale language model: refers to a large model trained using artificial intelligence technology.

[0035] LLM: Large-Scale Language Model, a technology that can generate high-quality text based on input text.

[0036] Knowledge document question and answer: refers to the use of artificial intelligence technology to process questions and answers in knowledge documents.

[0037] The specific embodiments of the present application are described in detail below in conjunction with the accompanying drawings.

[0038] Figure 2 A flow chart of a method for generating a reply based on a large-scale language model according to an exemplary embodiment of the present application is shown. Figure 2 The answer generation method shown includes steps S201, S203 and S205. Figure 2 , a reply generation method based on a large-scale language model according to an example embodiment of the present application is described in detail.

[0039] like Figure 2 As shown, in step S201, in response to a question instruction, it is determined whether the question included in the question instruction is an uncertain question.

[0040] In this application, the user question instruction input includes multiple forms, including but not limited to text input, voice input or any other form of input. The input question includes a format, including but not limited to text, picture, audio and / or the like.

[0041] In an embodiment of the present application, the question instruction includes a question input by the user. In step 201, according to the question input by the user, the intelligent customer service performs intent recognition and / or domain recognition on the input question to determine the intent or target of the text or voice input by the user.

[0042] When the user's intention can be accurately identified, the intelligent customer service is used to reply to the user. The intelligent customer service reply content includes multiple sources, for example, using the recall system to obtain the corresponding reply content from the knowledge base; or obtaining the corresponding reply content based on the search tool; or the intelligent customer service communicates with the merchant to obtain the corresponding reply content.

[0043] In a specific embodiment, the knowledge base includes, but is not limited to, a knowledge document question and answer knowledge base, a commodity knowledge base, a document knowledge base, and / or an extended knowledge base.

[0044] In some embodiments, before step S205, it is also necessary to train the intelligent customer service using a preset knowledge base so that the intelligent customer service can directly obtain corresponding answers or similar questions through vector recall based on the user's input questions, thereby improving the accuracy of task completion.

[0045] According to an embodiment of the present application, when the user's intention cannot be accurately identified, that is, when data still needs to be collected from the user, step S203 is executed.

[0046] In step S203, when it is determined that the problem is an uncertain problem, a task corresponding to the problem is generated.

[0047] According to an embodiment of the present application, before generating a task corresponding to a question, the input question needs to be preprocessed to convert the user's question into a preset data format, so that in step S203, the preset data format is used to generate a task corresponding to the question through a rule engine. The preset data format includes but is not limited to a dictionary format. When converting the user's question into the preset data format, the key value corresponding to the dictionary format is generated by extracting keywords from the user's question or retrieving the corresponding question description from the preset question description according to the user's question, and the preset dictionary format is filled to obtain the question description in the dictionary format.

[0048] In step S205, according to the generated task, an online conversation is conducted based on a large-scale language model to collect information corresponding to the task, thereby generating a response corresponding to the question.

[0049] According to an embodiment of the present application, in step S205, first, based on the generated task, a preset rule engine is used to determine the rules corresponding to the task; then, based on a large-scale language model, an information collection task is generated according to the rules; finally, at least one online conversation is conducted according to the generated information collection task to collect information corresponding to the task, thereby generating a reply corresponding to the question.

[0050] In a specific embodiment, first, according to the generated task, based on the order status corresponding to the question, the rules corresponding to the task are determined; then, the corresponding rules are executed, and the execution method includes but is not limited to having a conversation with the user, big store or supplier to collect the reply content corresponding to the question, or filling the slots of the preset reply format according to the reply corresponding to the question to generate the corresponding reply, or generating the reply corresponding to the question according to the collected information corresponding to the task and the preset knowledge base. According to an embodiment of the present application, the generated task corresponding to the question includes multiple tasks, including but not limited to using online conversations to collect question requirements or corresponding reply content; using a preset knowledge base to generate a reply corresponding to the question; and / or filling the preset reply template with content according to the input question to generate a reply corresponding to the input question.

[0051] According to an embodiment of the present application, Figure 2 The illustrated method also includes generating a response corresponding to the input question based on the response data collected by the plurality of tasks.

[0052] according to Figure 2 The embodiment shown converts the user's questions into tasks for online conversations to collect information corresponding to the tasks and generate replies corresponding to the questions, thereby improving the accuracy of intelligent customer service replies and alleviating the pressure on manual customer service.

[0053] In the process of generating a response corresponding to a question, when a single conversation cannot produce a definite response, according to other embodiments of the present application, multiple conversations are conducted with the user using the generated tasks, and steps S201 to S203 are executed based on the user input content, and this cycle is repeated multiple times until a definite response corresponding to the question is obtained.

[0054] Figure 3 A schematic diagram of an intelligent customer service workflow according to an exemplary embodiment of the present application is shown. Figure 3 As shown, the user selects one of multiple customer service entrances to input a question. The intelligent customer service first performs intent recognition based on the user input, and then selects the corresponding operation branch based on the intent recognition result.

[0055] For example, when the user's intention can be accurately identified, the customer service tool library is used to obtain the corresponding reply content from the knowledge base through the recall system. In some embodiments, the customer service tool library includes but is not limited to obtaining the same or similar replies from a preset knowledge base using a vector recall method, or obtaining the corresponding reply using a large-scale language model LLM.

[0056] For example, when the user's intention cannot be accurately identified, a large-scale language model (such as LLM) is used to communicate with the user and continue to identify the intention based on the user's response content.

[0057] For another example, when the user's intention cannot be accurately identified, first, a preset rule engine is used to generate multiple tasks, and then an online conversation is conducted using the customer service tool library based on the generated tasks to collect information corresponding to the tasks, or a reply corresponding to the question is directly generated based on the generated tasks or directly sent to a manual customer service application for manual processing.

[0058] according to Figure 3 The illustrated embodiment abstracts the conditions that need to be collected into multiple tasks, and uses a large-scale language model LLM to complete the tasks. By introducing a rule engine, certainty is brought to the uncertainty in the LLM task execution.

[0059] In some embodiments, after the LLM task is completed, the entire process can be recursively repeated to complete the collection of user requirements through multiple sessions.

[0060] Figure 4A schematic diagram of a workflow of a rule engine according to an exemplary embodiment of the present application is shown. Figure 4 As shown in the figure, firstly, the intention is recognized according to the user's question input, and the user's question is converted into a preset data format. The intention recognition result is as follows Figure 5 In a specific embodiment, intent recognition includes retrieval rewriting, domain recognition and intent recognition.

[0061] Based on intent recognition and order status information corresponding to user questions (such as Figure 6 As shown), a corresponding task is generated through a preset rule engine, and the corresponding rule is executed according to the generated task type and the rule configuration of the rule engine.

[0062] For example, if the generated task is to collect user requirements, then the rule corresponding to the task is to conduct a conversation with the user to collect requirements.

[0063] For another example, if the generated task is to get answers from merchants, the rule corresponding to the task is to have a conversation with suppliers or large store customer service to get the corresponding answers to questions. In some embodiments, conversation tasks with different boundaries are generated according to the preset slot filling definition. In this embodiment, although the logic requirements of the dialogue rules written in the traditional way are not as high, the conversation strategy with more comprehensive logical branches in this embodiment can better guide the large-scale language model LLM to have online conversations with users and obtain more accurate user needs.

[0064] For another example, the generated task is to fill in the reply content based on the intent recognition result, such as Figure 7 As shown, the rule corresponding to the task is to generate corresponding question replies to the customer.

[0065] According to an embodiment of the present application, the generated task also includes collecting output data of various tasks to generate answers to questions.

[0066] In an embodiment of the present application, the generated task also includes confirming with the user whether the problem has been solved. When the problem has not been solved, in response to the user's continue execution instruction, the conversation with the user is continued, intent recognition is performed, corresponding tasks are generated to obtain answers to the questions, and the like are repeated multiple times until the customer requirements are met.

[0067] according to Figure 4 In the embodiment shown, although the complexity of the problems that the intelligent customer service needs to solve ranges from the simplest product answers to complex after-sales negotiations and follow-ups, in this embodiment, the user's input questions are uniformly and concisely abstracted and corresponding tasks are generated, so that the same LLM tool can be used to generate answers corresponding to the input questions.

[0068] According to the embodiments of the present application, there is no need to hand-write complex conversation rules or a large amount of high-quality labeled conversation data, and the development cost is low. And according to the preset task goals, slot values ​​and related contexts, there is still a strong multi-round conversation management capability for users' unconventional paths. And according to the generated multiple task types, it is possible to get more accurate answers corresponding to the input questions.

[0069] The above mainly introduces the embodiments of the present application from the perspective of the method. Those skilled in the art should easily appreciate that, in combination with the operations or steps of each example described in the embodiments disclosed herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Those skilled in the art can use different ways to implement the described functions for each specific operation or method, and such implementation should not be considered to be beyond the scope of the present application.

[0070] The following describes the device embodiments of the present application. For details not described in the device embodiments of the present application, reference can be made to the method embodiments of the present application.

[0071] Figure 8 A block diagram of a reply generation device based on a large-scale language model according to an exemplary embodiment of the present application is shown, Figure 8 The answer generation device shown includes a question feature determination unit 801, a task generation unit 803 and an answer generation unit 805. The question feature determination unit 801 is used to determine whether the question included in the question instruction is an uncertainty question in response to the question instruction; the task generation unit 803 is used to generate a task corresponding to the question when the question is determined to be an uncertainty question; the answer generation unit 805 is used to conduct an online conversation based on the generated task and the large-scale language model to collect information corresponding to the task, thereby generating an answer corresponding to the question.

[0072] Fig. 9 An electronic device according to an exemplary embodiment of the present application is shown. Fig. 9 The electronic device 200 according to this embodiment of the present application is described. Fig. 9 The electronic device 200 shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0073] like Fig. 9 As shown, the electronic device 200 is in the form of a general computing device. The components of the electronic device 200 may include, but are not limited to: at least one processing unit 210, at least one storage unit 220, a bus 230 connecting different system components (including the storage unit 220 and the processing unit 210), a display unit 240, etc.

[0074] The storage unit stores program codes, which can be executed by the processing unit 210, so that the processing unit 210 executes the methods described in this specification according to various exemplary embodiments of the present application. For example, the processing unit 210 can execute the following: Figure 1 The method shown in .

[0075] The storage unit 220 may include a readable medium in the form of a volatile storage unit, such as a random access memory unit (RAM) 2201 and / or a cache memory unit 2202 , and may further include a read-only memory unit (ROM) 2203 .

[0076] The storage unit 220 may also include a program / utility 2204 having a set (at least one) of program modules 2205, such program modules 2205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0077] Bus 230 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0078] The electronic device 200 may also communicate with one or more external devices 300 (e.g., keyboards, pointing devices, Bluetooth devices, etc.), may also communicate with one or more devices that enable a user to interact with the electronic device 200, and / or communicate with any device that enables the electronic device 200 to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed via an input / output (I / O) interface 250. Furthermore, the electronic device 200 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 260. The network adapter 260 may communicate with other modules of the electronic device 200 via the bus 230. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 200, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0079] Through the description of the above implementation methods, it is easy for those skilled in the art to understand that the example implementation methods described here can be implemented by software or by combining software with necessary hardware. The technical solution according to the implementation method of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, or a network device, etc.) to execute the above method according to the implementation method of the present application.

[0080] The software product may use any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0081] Computer readable storage media may include data signals propagated in baseband or as part of a carrier wave, wherein readable program codes are carried. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The readable storage medium may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, device, or device. The program codes contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination thereof.

[0082] Program code for performing the operations of the present application may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0083] The computer-readable medium carries one or more programs. When the one or more programs are executed by a device, the computer-readable medium implements the aforementioned functions.

[0084] Those skilled in the art will appreciate that the above modules can be distributed in the device according to the description of the embodiment, or can be changed accordingly and only used in one or more devices different from the embodiment. The modules of the above embodiments can be combined into one module, or further divided into multiple sub-modules.

[0085] According to an embodiment of the present application, a computer program is provided, including a computer program or an instruction. When the computer program or the instruction is executed by a processor, the method described above can be executed.

[0086] The embodiments of the present application are described in detail above. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and its core idea of ​​the present application. At the same time, changes or deformations made by those skilled in the art based on the ideas of the present application, the specific implementation methods and the scope of application of the present application, all belong to the scope of protection of the present application. In summary, the content of this specification should not be construed as a limitation on the present application.

Claims

1. A reply generation method based on a large-scale language model, characterized in that: include: In response to a question instruction, determining whether the question included in the question instruction is an uncertain question; In the case where it is determined that the problem is an uncertain problem, generating a task corresponding to the problem; According to the generated task, an online conversation is conducted based on the large-scale language model to collect information corresponding to the task, thereby generating a response corresponding to the question.

2. The answer generation method according to claim 1, characterized in that: The question instruction includes a question input by a user, and determining whether the question included in the question instruction is an uncertain question includes: Intention recognition is performed based on the input question to determine whether the question included in the question instruction is an uncertain question.

3. The answer generation method according to claim 1, characterized in that: Generate tasks corresponding to the problem, including: Using a preset data format, a task corresponding to the problem is generated.

4. The answer generation method according to claim 3, characterized in that: According to the generated task, based on the large-scale language model, an online conversation is conducted to collect information corresponding to the task, thereby generating a reply corresponding to the question, including: According to the task, using a preset rule engine, determining a rule corresponding to the task; Based on the large-scale language model, generating an information collection task according to the rules; Conduct at least one online conversation according to the generated information collection task to collect information corresponding to the task, thereby generating a response corresponding to the question.

5. The answer generation method according to claim 2, characterized in that: Generating tasks corresponding to the above problems includes: Use online conversations to collect questions or corresponding responses; Generate a response corresponding to the question using a preset knowledge base; and / or Based on the input question, the preset reply template is filled with content to generate a reply corresponding to the input question.

6. The answer generation method according to claim 5, characterized in that: The generated tasks corresponding to the question include a plurality of tasks, and the answer generation method further includes: A response corresponding to the input question is generated based on the response data collected by the plurality of tasks.

7. The answer generation method according to claim 1, characterized in that: Also includes: Conducting an online conversation with the user to determine whether the user's problem has been resolved; In response to the user's instruction to continue executing, the system continues to generate a reply corresponding to the new question instruction input by the user again, until the user's instruction to end the session is received.

8. A reply generation device based on a large-scale language model, characterized in that: include: a question feature determination unit, configured to determine, in response to a question instruction, whether the question included in the question instruction is an uncertain question; A task generating unit, configured to generate a task corresponding to the problem when determining that the problem is an uncertain problem; The answer generation unit is used to conduct an online conversation based on the generated task and the large-scale language model to collect information corresponding to the task, thereby generating an answer corresponding to the question.

9. An electronic device, characterized in that: include: processor; Memory for storing computer programs; When the computer program is executed by the processor, the processor is caused to implement the method according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having computer-readable instructions stored thereon, which, when executed by a processor, causes the processor to execute the method according to any one of claims 1 to 7.

Images