Call-out management method and device based on large language model, electronic equipment, storage medium and computer product

Abstract

The invention relates to the technical field of communication, and provides an outbound call management method and device based on a large language model, electronic equipment, a storage medium and a computer product, and the method comprises the steps: obtaining dialogue information returned after an outbound call; inputting the dialogue information into a question and answer model, and if an answer text corresponding to the dialogue information does not exist in output information of the question and answer model, inputting the dialogue information into a preset large language model to obtain an answer text output by the preset large language model; wherein the question and answer model is used for performing answer retrieval based on input dialogue information; preset cue words are set in the preset large language model; the preset prompt word is generated based on outbound process circulation information and an outbound service type response rule; and outputting the answer based on the answer text. According to the method and the device, accurate outbound response can be realized without tedious process configuration, and the outbound efficiency can be improved.

Description

Technical Field

[0001] This application relates to the field of communication technologies, and particularly to an outbound call management method, apparatus, electronic device, storage medium, and computer product based on a large language model. Background Art

[0002] Intelligent outbound call systems are widely used in many industries and business scenarios, which can significantly improve work efficiency and customer satisfaction. Currently, common implementation solutions include underlying natural language processing (NLP) technologies, combined with advanced speech technologies such as automatic speech recognition (ASR) and text-to-speech (TTS). These technologies work together to enable the system to understand and generate natural language, and achieve intelligent conversations with customers.

[0003] An intelligent outbound call system usually defines and manages the outbound call process through a process canvas. The process canvas includes the following main elements: Node information: including start nodes, reply nodes, multi-round dialogue nodes, assignment nodes, and hang-up nodes, etc. These nodes define each step and operation of the outbound call process.

[0004] Trigger lines: The trigger lines between nodes are used to connect different nodes and determine the flow direction of the process according to the system logic and user input.

[0005] Intent recognition: Configure different intents, such as affirmative, negative, busy, etc. intents, to judge the user's answer and select the corresponding trigger line according to the intent.

[0006] Global variables: Manage global variables, including input parameters and output parameters, to transfer and share data between different nodes, ensuring the coherence and consistency of the process.

[0007] In actual operation, users can perform batch outbound calls by creating tasks. This includes defining a target customer list, setting the outbound call time and strategies (such as the number of calls and the interval time), and configuring the specific outbound call process. The system will automatically make outbound calls according to the preset task arrangement, and interact with customers through NLP and speech technologies to collect feedback or provide services. However, traditional outbound call solutions usually have the problem of cumbersome process configuration, which in turn leads to low efficiency when making outbound calls currently. Summary of the Invention

[0008] This application aims to solve at least one of the technical problems existing in the related art. To this end, this application proposes an outbound call management method, device, electronic device, storage medium and computer product based on a large language model to solve the problem of cumbersome process configuration in traditional outbound call solutions and achieve improved outbound call efficiency.

[0009] The outbound call management method based on a large language model according to the first aspect embodiment of this application includes: Obtain the conversation information returned after the outbound call; Input the conversation information into the Q&A model. If the answer text corresponding to the conversation information does not exist in the output information of the Q&A model, input the conversation information into the preset large language model to obtain the answer text output by the preset large language model; wherein, the Q&A model is used to perform answer retrieval based on the input conversation information; preset prompt words are set in the preset large language model; the preset prompt words are generated based on the answer rules of the process flow information of the outbound call and the business type of the outbound call; Output the answer based on the answer text.

[0010] According to an embodiment of this application, the outputting the answer based on the answer text includes: Determine the conversation language based on the conversation information; Optimize the answer text based on the conversation language to obtain the target text; Convert the target text into audio and then output it.

[0011] According to an embodiment of this application, the Q&A model further includes Q&A pairs, and the Q&A model is further used to perform answer retrieval of the input conversation information based on the Q&A pairs by using a text similarity algorithm.

[0012] According to an embodiment of this application, the Q&A model is further used to call a third-party tool based on the conversation information to determine the output information.

[0013] According to an embodiment of this application, the preset large language model is further used to cache the historical conversation information of the outbound call user.

[0014] According to an embodiment of this application, after inputting the conversation information into the Q&A model, it further includes: If the output information of the Q&A model includes the answer text corresponding to the conversation information, output the answer based on the answer text.

[0015] The outbound call management device based on a large language model according to the second aspect embodiment of this application includes: An acquisition module, configured to acquire the conversation information returned after the outbound call; An input module for inputting the conversation information into a question-and-answer model. If there is no answer text corresponding to the conversation information in the output information of the question-and-answer model, the conversation information is input into a preset large language model to obtain the answer text output by the preset large language model. Among them, the question-and-answer model is used to retrieve answers based on the input conversation information; there is a preset prompt word in the preset large language model; the preset prompt word is generated based on the process flow information of the outbound call and the response rules of the outbound call business type. An output module for outputting answers based on the answer text.

[0016] An electronic device according to an embodiment of the third aspect of the present application includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements the outbound call management method based on a large language model as described in any one of the above.

[0017] According to a storage medium of an embodiment of the fourth aspect of the present application, the storage medium is a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the outbound call management method based on a large language model as described in any one of the above.

[0018] A computer program product according to an embodiment of the fifth aspect of the present application includes a computer program. When the computer program is executed by a processor, it implements the outbound call management method based on a large language model as described in any one of the above.

[0019] One or more of the above technical solutions in the embodiments of the present application have at least the following technical effects: By formulating corresponding response rules in advance based on the outbound call business type, it is possible to generate a preset prompt word according to the process flow information of the outbound call and the response rules and set it in the preset large language model. Furthermore, when the question-and-answer model used to retrieve answers based on the input conversation information cannot obtain the answer to the conversation information returned after the outbound call, the answer text can be queried by the method of combining the large language model with the prompt word, and finally the answer is output based on the answer text to complete the outbound call task. Since only the process flow information of the outbound call needs to be set, the outbound call task can be accurately completed in combination with the response rules formulated based on the outbound call business type, and accurate outbound call responses can be achieved without cumbersome process configuration, which can improve the outbound call efficiency.

[0020] Additional aspects and advantages of the present application will be given in part in the following description, become apparent in part from the following description, or be learned through the practice of the present application. Description of the Drawings

[0021] To more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0022] Figure 1 It is a schematic flowchart of the outbound call management method based on a large language model provided by an embodiment of this application.

[0023] Figure 2 It is a schematic structural diagram of an electronic device provided by this application. Specific embodiments

[0024] The following further describes the embodiments of this application in detail with reference to the drawings and embodiments. The following embodiments are used to illustrate this application, but cannot be used to limit the scope of this application.

[0025] In the description of the embodiments of this application, it should be noted that the orientation or positional relationship indicated by the terms "center", "longitudinal", "transverse", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the embodiments of this application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the embodiments of this application. In addition, the terms "first", "second", and "third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0026] In the description of the embodiments of this application, it should be noted that unless otherwise clearly specified and limited, the terms "connected" and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium. For those of ordinary skill in the art, the specific meanings of the above terms in the embodiments of this application can be understood according to specific situations.

[0027] In the embodiments of the present application, unless otherwise clearly specified and limited, the first feature being "on" or "under" the second feature may mean that the first and second features are in direct contact, or the first and second features are indirectly in contact through an intermediate medium. Moreover, the first feature being "above", "over" and "on top of" the second feature may mean that the first feature is directly above or obliquely above the second feature, or simply means that the horizontal height of the first feature is higher than that of the second feature. The first feature being "under", "beneath" and "underneath" the second feature may mean that the first feature is directly below or obliquely below the second feature, or simply means that the horizontal height of the first feature is less than that of the second feature.

[0028] In the description of this specification, the description with reference to terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the embodiments of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0029] The present application provides an outbound call management method, device, electronic device, storage medium, and computer product based on a large language model.

[0030] Figure 1 is a schematic flowchart of the outbound call management method based on a large language model provided by the embodiments of the present application, as Figure 1 shown, the outbound call management method based on a large language model includes: Step 110, obtaining the conversation information returned after the outbound call.

[0031] Step 120, inputting the conversation information into the Q&A model. If the answer text corresponding to the conversation information does not exist in the output information of the Q&A model, input the conversation information into the preset large language model to obtain the answer text output by the preset large language model; wherein, the Q&A model is used to retrieve answers based on the input conversation information; preset prompt words are set in the preset large language model; the preset prompt words are generated based on the answer rules of the process flow information of the outbound call and the business type of the outbound call.

[0032] Step 130, performing answer output based on the answer text.

[0033] It should be noted that the execution subject of the outbound call management method based on the large language model provided in the embodiments of the present application can be a server, a computer device, etc. The computer device can be, for example, a mobile phone, a tablet computer, a laptop computer, a handheld computer, an in-vehicle electronic device, a wearable device, an Ultra-mobile Personal Computer (UMPC), a netbook, or a Personal Digital Assistant (PDA), etc. It should be noted that all the data to be obtained in the present application is obtained through regular channels after being authorized by relevant users.

[0034] In the server and computer device of the present application, an outbound call management device based on the large language model can be set or connected, so that the outbound call management device based on the large language model can be controlled to execute the outbound call management method based on the large language model of the present application.

[0035] The present application can pre-construct and train a model that can retrieve answers according to the input text as a question-and-answer model.

[0036] At the same time, some special markers can be pre-defined. For example, the first command "start" represents that the user is ready to start a conversation, "silence" represents that the user is silent, and "#end#" represents that the call needs to be terminated. Secondly, in terms of process interaction, combining the experience accumulated in the traditional process, highly abstract the common scenario processing (such as scenarios where the user is busy, the elderly cannot operate, a child answers the phone, identity inquiry, teasing, the user refuses to return the call, the user abuses and is unwilling to participate, etc.). During the interaction process, the user may consult questions related to sensitive words such as porn, gambling, and drugs. The relevant content of the reply is strictly restricted through prompt words. In addition, when the user consults some irrelevant questions, prompt words are configured to guide the user back to the process. Some emotion perception constraints are configured to perceive the user's emotions such as joy and anger and generate corresponding replies. Thus, the answer rules for outbound calls under various business types are obtained as prompt word templates. Relevant personnel only need to describe information such as the role, goal, questions to be collected, intention, relevant detailed indicators, start reply, and end reply related to the process as the process flow information for outbound calls.

[0037] It should be noted that the process flow information may also include key information extraction prompts. Specifically, this application can indicate that the collection of some user information has completed the return visit measurement. The traditional approach is to trigger the intention of certain users and then assign values to statistical values. Since it is local recognition, it cannot well understand the intention that the user wants to express. This application uses a large language model to read the conversation content to summarize the conversation content. In addition, the user can customize the intention rules, such as: 1. Participate in the return visit, 2. Have feedback on problems, 3. The user is busy, 4. Resist the return visit, 5. Unwilling to participate in the return visit. The intention degree rules are described as follows: Participate in the return visit and rate: Users who give ratings are classified into this category.

[0038] Have feedback on problems: Have feedback on problems.

[0039] The user is busy: The user indicates being busy.

[0040] Resist the return visit: The user abuses.

[0041] Unwilling to participate in the return visit: **Users without conversation records; calls less than 10 seconds; or those not meeting the above conditions are classified as unwilling to participate in the return visit**

[0042] If there are questions to be collected, the large language model will extract the answers to the questions.

[0043] Therefore, after the conversation ends, by using the powerful reasoning ability of the large language model, it is possible to summarize the complete call and extract key information, significantly improving the processing efficiency.

[0044] Furthermore, for outbound calls of different business types, relevant personnel can obtain the corresponding response rules for the business types and, together with the process flow information of the outbound calls, set them as preset prompts in the existing large language model to obtain a preset large language model. Among them, the large language model generally refers to a deep learning model with a large number of parameters and powerful capabilities, especially in the field of natural language processing. These models can handle complex language tasks, such as text generation, question-and-answer systems, language understanding, etc.

[0045] By using natural language descriptions to replace the traditional cumbersome process configuration, making full use of the excellent ability of the large language model in natural language processing, it overcomes the limitations of traditional NLP that rely on prefabricated intentions and fixed script responses.

[0046] Large language models can include but are not limited to the GPT series, such as GPT-4 and GPT-4o, which perform excellently in multiple fields such as language understanding, generation, and code writing, demonstrating powerful natural language processing capabilities. The BERT series, including the BERT model and its subsequent improved versions such as RoBERTa, ALBERT, etc., has made significant progress in multiple NLP tasks through pre-training techniques. And, other existing large language models.

[0047] By combining large language models with prompt words, responses can be flexibly generated according to the context, reducing the dependence on preset intents.

[0048] Furthermore, a preset large language model and a question-and-answer model can be integrated into an outbound call system or platform together, and it can be set in the outbound call system or platform (subsequently, the outbound call platform can be taken as an example for illustration) that when an answer cannot be obtained through the question-and-answer model, the preset large language model is called to retrieve the answer.

[0049] It should be further noted that common intents can also be built into the outbound call platform of this application for testing. The intents can include, for example, children, the elderly, being busy, affirmative intent, negative intent, identity inquiry, hearing impairment, robot, complaint, abuse, off-topic answers, etc.

[0050] And, according to the user's process information, existing intelligent models can be used to automatically generate scenarios and corresponding sets of conversation scripts corresponding to various intents. Then, tests can be conducted based on these test sets, reducing the limitations of manual testing and solving the problem of the cumbersome traditional supplementary data.

[0051] When both the question-and-answer model and the preset large language model pass the tests, the outbound call platform can be officially enabled for application.

[0052] This application conducts tests by building in conventional scenarios, solving the cumbersome step-by-step testing process. And, through intent and scenario dialogue testing, the accuracy and effectiveness of the outbound call platform are ensured, reducing problems after going live.

[0053] Evaluating the dialogue effect using common scenario test sets also effectively reduces the steps of making individual phone calls for each branch in traditional testing, significantly reducing the testing cost and further improving production efficiency.

[0054] Thus, the outbound call platform can make an outbound call to a specified user according to the outbound call instruction, obtain the user's conversation voice after connection, and then convert the conversation voice into conversation text through automatic speech recognition technology and determine it as conversation information.

[0055] Furthermore, the conversation information can be input into the Q&A model. Among them, the Q&A model can include Q&A pairs. Specifically, this application can systematically cache common Q&A pairs into the Q&A model, especially in scenarios where the user indicates that they are busy, and preset some common expressions. These expressions include: "I'm busy", "I'm driving", "I don't have time at work", "I'm cooking", etc. For example, when the user says "Well, I'm driving", through text similarity, it can quickly recognize that their intention matches the preset statement "I'm driving", so that the corresponding reply can be quickly found and returned. This process can significantly reduce the response time and improve the user experience.

[0056] In addition, after the program of the outbound calling platform runs for a period of time, high-quality Q&A pairs generated by the preset large language model can be cached and saved to gradually enrich the corpus of the model. This accumulation can not only improve the response efficiency of the outbound calling platform to common questions, but also further optimize the preset Q&A pairs by analyzing and summarizing the user's question patterns. This means that over time, the number of requests sent to the large model will gradually decrease, thereby reducing the operating burden of the system.

[0057] Through such a design, it is possible to quickly reuse these stored Q&A pairs and perform generalization processing on the user's intention in combination with the text similarity model. This method can not only speed up the response speed to common questions, but also improve the overall performance of the system, ensuring that the user experience in the interaction is smoother and more efficient. In short, this application aims to provide a more user-friendly service while improving efficiency through intelligent Q&A matching and effective data accumulation.

[0058] Therefore, the Q&A model can retrieve the answer to the input conversation information using the text similarity algorithm based on the Q&A pairs and the existing knowledge base, and determine the output information according to the retrieval result.

[0059] It should be noted that the Q&A model of this application can also call third-party tools based on the conversation information to determine the output information. Specifically, during the process interaction, third-party tools may need to be triggered according to the conversation information, such as specific tools, interfaces, custom program methods, etc. to assist in completing the process interaction. More specifically, the third-party tools can include but are not limited to sending text messages, transferring to a human operator, date tools, etc.

[0060] Specifically, for the call of third-party tools, it can be called according to the corresponding trigger instructions. For example, if the user's conversation information is "What day of the week is today?", the date tool can be triggered to determine the specific day of the week, and the corresponding date can be combined with other information that needs to be returned to generate the output information.

[0061] By integrating third-party tools (search, send text messages, transfer to human, etc.) and custom code tools, the limitations of traditional process tool sets are solved.

[0062] Furthermore, if the answer text corresponding to the conversation information does not exist in the output information of the Q&A model, a preset large model can be called, the conversation information can be input into the preset large language model, and the preset large language model can retrieve the answer text corresponding to the conversation information according to the preset prompt words. After the preset large language model completes the retrieval, the answer text output by the preset large language model can be obtained.

[0063] It should be noted that the historical conversation information of the outbound user can also be cached in the preset large language model of the present application.

[0064] Specifically, traditional dialogue systems lack context understanding and coherence when processing multi-round conversations, resulting in users needing to repeat information, being unable to obtain a personalized experience, and performing poorly in complex interactions. The present application caches the historical conversation information of the outbound user in the preset large language model and the Q&A model by using redis caching technology, thereby recording the content of the human-machine conversation, enabling the model to have memory ability, and making the conversation more fluent and intelligent. In an interaction case, the user said, "I didn't hear clearly. Say it again." The model retrieved the cached content and replied, "How many points do you give for the overall performance of China Mobile's mobile phone tariffs, network quality, etc.?" This makes the conversation connection more natural. Among them, redis is an open-source high-performance key-value database, which is usually used as a database, cache, and message passing system.

[0065] The present application solves the problem that traditional process local conversations need to consult the user's information again to reply by caching the memory in the storage medium, endowing the model with memory ability, and at the same time providing a data basis for information extraction, improving the satisfaction of customers during the conversation.

[0066] Furthermore, the answer text can be converted into speech and output to the user. After all the user's questions are answered and it is detected that an instruction to terminate the current call is required, the current outbound call is ended, thereby completing the outbound call task.

[0067] The present application summarizes and extracts key information from the conversation by describing key indicators such as roles, goals, and questions in natural language and combining the natural language processing and conversation understanding capabilities of the large language model. This can improve the information processing efficiency, eliminate the dependence on traditional process configurations, and give full play to the advantages of the large language model in understanding and reasoning.

[0068] According to the outbound call management method based on a large language model in the embodiments of the present application, by formulating corresponding response rules in advance based on the business types of outbound calls, it is possible to generate preset prompt words according to the process flow information of outbound calls and the response rules and set them in a preset large language model. Furthermore, when the answer to the dialogue information returned after an outbound call cannot be obtained through a question-and-answer model used to retrieve answers based on the input dialogue information, the answer text can be queried by means of the large language model plus the prompt words, and finally the answer is output based on the answer text to complete the outbound call task. Since only the process flow information of the outbound call needs to be set, the outbound call task can be accurately completed in combination with the response rules formulated based on the business types of outbound calls, and accurate outbound call responses can be achieved without cumbersome process configuration, which can improve the outbound call efficiency.

[0069] Based on the above embodiments, the answer output based on the answer text includes: Determine the dialogue language based on the dialogue information; Optimize the answer text based on the dialogue language to obtain the target text; Convert the target text into audio and then output it.

[0070] It should be noted that existing intelligent dialogue systems are often restricted by the single language of the conversation terms when dealing with multiple languages or dialects, and it is difficult to provide a personalized and realistic dialogue experience.

[0071] To overcome this limitation, the present application can determine the dialogue language based on the dialogue information or dialogue voice. For example, specifically, the corresponding voice or text can be determined as to which language it belongs to through speech recognition technology. In one embodiment, English, Mandarin, Cantonese, etc. can be recognized.

[0072] Furthermore, optimizing the answer text according to the corresponding dialogue language can specifically be to optimize the original answer text into the text of the corresponding language sentence pattern as the target text. For example, adjusting the original answer text in Mandarin to an English text as the target text.

[0073] Furthermore, through text-to-speech technology, using a hyper-realistic text synthesis voice, the target text is converted into audio and output, making the output content more intelligent and the voice more similar to that of a real person, thereby enhancing the authenticity of the dialogue and the customer's trust.

[0074] The present application solves the problem of the fixed and single language of the traditional process reply by identifying the user's language and dynamically switching relevant languages for reply, significantly improving the user experience in a multi-language environment, enhancing the communication efficiency, and strengthening the effect of cross-language communication.

[0075] Based on the above embodiments, after inputting the dialogue information into the question-and-answer model, it further includes: If the output information of the Q&A model includes the answer text corresponding to the conversation information, the answer is output based on the answer text.

[0076] Specifically, if the output information of the Q&A model includes the answer text corresponding to the conversation information, since the answer has been retrieved, the answer can be directly output.

[0077] More specifically, the conversation language can be determined based on the conversation information or conversation voice. For example, specifically, speech recognition technology can be used to determine which language the corresponding voice or text belongs to. In one embodiment, English, Mandarin, Cantonese, etc. can be recognized.

[0078] Furthermore, the answer text is optimized according to the corresponding conversation language. Specifically, the original answer text can be optimized into a text in the corresponding language sentence pattern as the target text. For example, the original answer text in Mandarin is adjusted to an English text as the target text.

[0079] When the present application can retrieve an answer through the Q&A model, the corresponding answer can be directly output, thereby improving the communication efficiency of the outbound call.

[0080] Next, the outbound call management device based on the large language model provided by the present application is described. The outbound call management device based on the large language model described below can be mutually referred to the outbound call management method based on the large language model described above.

[0081] Furthermore, the present application also provides an outbound call management device based on the large language model.

[0082] The outbound call management device based on the large language model includes: An acquisition module, configured to acquire the conversation information returned after the outbound call; An input module, configured to input the conversation information into the Q&A model. If the answer text corresponding to the conversation information does not exist in the output information of the Q&A model, the conversation information is input into the preset large language model to obtain the answer text output by the preset large language model; wherein, the Q&A model is used to retrieve an answer based on the input conversation information; preset prompt words are set in the preset large language model; the preset prompt words are generated based on the answer rules of the process flow information of the outbound call and the business type of the outbound call; An output module, configured to output an answer based on the answer text.

[0083] The outbound call management device based on the large language model of the present application formulates corresponding response rules in advance based on the business type of the outbound call, so that preset prompt words can be generated according to the process flow information of the outbound call and the response rules and set in the preset large language model. Furthermore, when the question-and-answer model used to retrieve answers based on the input dialogue information cannot obtain the answer to the dialogue information returned after the outbound call, the answer text can be queried by means of the large language model plus the prompt words, and finally the answer is output based on the answer text to complete the outbound call task. Since only the process flow information of the outbound call needs to be set, the outbound call task can be accurately completed in combination with the response rules formulated based on the business type of the outbound call, and accurate outbound call responses can be achieved without cumbersome process configuration, which can improve the outbound call efficiency.

[0084] In one embodiment, the input module is further configured to: If the output information of the question-and-answer model includes the answer text corresponding to the dialogue information, output the answer based on the answer text.

[0085] In one embodiment, the output module is specifically configured to: Determine the dialogue language based on the dialogue information; Optimize the answer text based on the dialogue language to obtain the target text; Convert the target text into audio and then output it.

[0086] Figure 2 An example of the physical structure diagram of an electronic device is shown as Figure 2 shown. The electronic device may include: a processor 210, a communication interface 220, a memory 230, and a communication bus 240. Among them, the processor 210, the communication interface 220, and the memory 230 communicate with each other through the communication bus 240. The processor 210 can call the logical instructions in the memory 230 to execute the following methods: obtain the dialogue information returned after the outbound call; Input the dialogue information into the question-and-answer model. If the output information of the question-and-answer model does not include the answer text corresponding to the dialogue information, input the dialogue information into the preset large language model to obtain the answer text output by the preset large language model; wherein, the question-and-answer model is used to retrieve answers based on the input dialogue information; preset prompt words are set in the preset large language model; the preset prompt words are generated based on the process flow information of the outbound call and the response rules of the business type of the outbound call; Output the answer based on the answer text.

[0087] In addition, when the logical instructions in the above-mentioned memory 230 can be implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the related technology, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0088] In another aspect, an embodiment of the present application further provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is configured to execute the methods provided in the above-mentioned various embodiments, for example, including: obtaining the conversation information returned after an outbound call; Inputting the conversation information into a question-and-answer model. If there is no answer text corresponding to the conversation information in the output information of the question-and-answer model, inputting the conversation information into a preset large language model to obtain the answer text output by the preset large language model; wherein, the question-and-answer model is used to retrieve answers based on the input conversation information; a preset prompt word is set in the preset large language model; the preset prompt word is generated based on the process flow information of the outbound call and the response rules of the business type of the outbound call; Outputting an answer based on the answer text.

[0089] In another aspect, an embodiment of the present application further provides a computer program product, on which a computer program is stored. When the computer program is executed by a processor, it is configured to execute the methods provided in the above-mentioned various embodiments, for example, including: obtaining the conversation information returned after an outbound call; Inputting the conversation information into a question-and-answer model. If there is no answer text corresponding to the conversation information in the output information of the question-and-answer model, inputting the conversation information into a preset large language model to obtain the answer text output by the preset large language model; wherein, the question-and-answer model is used to retrieve answers based on the input conversation information; a preset prompt word is set in the preset large language model; the preset prompt word is generated based on the process flow information of the outbound call and the response rules of the business type of the outbound call; Outputting an answer based on the answer text.

[0090] The device embodiments described above are merely illustrative. 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 to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative effort.

[0091] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0092] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features. And these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An outbound call management method based on a large language model, characterized in that: include: Get the conversation information returned after the outbound call; The dialogue information is input into a question-answering model. If the answer text corresponding to the dialogue information does not exist in the output information of the question-answering model, the dialogue information is input into a preset large language model to obtain the answer text output by the preset large language model; wherein the question-answering model is used to retrieve answers based on the input dialogue information; the preset large language model is provided with preset prompt words; the preset prompt words are generated based on the process flow information of the outbound call and the answering rules of the service type of the outbound call; An answer is output based on the answer text.

2. The outbound call management method based on a large language model according to claim 1, characterized in that: The outputting of the answer based on the answer text comprises: determining a conversation language based on the conversation information; Performing text optimization on the answer text based on the dialogue language to obtain a target text; The target text is converted into audio and then output.

3. The outbound call management method based on a large language model according to claim 1, characterized in that: The question-answer model also includes question-answer pairs, and the question-answer model is also used to retrieve answers to input dialogue information based on the question-answer pairs using a text similarity algorithm.

4. The outbound call management method based on a large language model according to claim 1, characterized in that: The question-answering model is also used to call a third-party tool based on the dialogue information to determine output information.

5. The outbound call management method based on a large language model according to claim 1, characterized in that: The preset large language model is also used to cache historical conversation information of outbound call users.

6. The outbound call management method based on a large language model according to claim 1, characterized in that: After inputting the dialogue information into the question-answering model, the method further includes: If the output information of the question-answer model includes an answer text corresponding to the dialogue information, an answer is output based on the answer text.

7. An outbound call management device based on a large language model, characterized in that: include: The acquisition module is used to obtain the conversation information returned after the outbound call; An input module is used to input the dialogue information into a question-answering model. If the answer text corresponding to the dialogue information does not exist in the output information of the question-answering model, the dialogue information is input into a preset large language model to obtain the answer text output by the preset large language model; wherein the question-answering model is used to retrieve answers based on the input dialogue information; the preset large language model is provided with preset prompt words; the preset prompt words are generated based on the process flow information of the outbound call and the answering rules of the service type of the outbound call; An output module is used to output an answer based on the answer text.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the outbound call management method based on the large language model as described in any one of claims 1 to 6 is implemented.

9. A storage medium, the storage medium being a non-transitory computer-readable storage medium, on which a computer program is stored, characterized in that: When the computer program is executed by a processor, the outbound call management method based on a large language model as described in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the outbound call management method based on a large language model described in any one of claims 1 to 6 is implemented.

Images