LLM-based consultation record summarization method, apparatus and device, medium and product

Abstract

The invention discloses an LLM-based consultation record summarization method and device, equipment, a medium and a product, and relates to the technical field of data processing, and the method comprises the steps: collecting multi-source customer service data; wherein the multi-source customer service data at least comprises session record data, call record data and work order information; preprocessing the multi-source customer service data to obtain target multi-source customer service data; processing the target multi-source customer service data by using a large language model to obtain initial summary content; and based on the initial summary content and a preset standardized template, generating standardized consultation record summary content, thereby effectively reducing possible negligence and deviation of manual summary, and greatly improving the accuracy of consultation record summary.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a consultation record summarization method and device based on a large language model (LLM), equipment, medium and product. BACKGROUND

[0002] A customer service system generates a large amount of consultation records every day. These records contain key information such as customer problems and needs, and are important basis for enterprises to optimize services and improve satisfaction. It is crucial to summarize consultation records in a timely and accurate manner.

[0003] However, the current consultation record summary of the customer service system mainly relies on manual work. This method is inefficient and the quality is not guaranteed. In the face of complex consultation records, customer service personnel may miss important details due to negligence or fatigue, and these details are crucial to solving problems, improving services and discovering business opportunities. Missing these details can cause losses to the enterprise. Therefore, the accuracy of the existing consultation record summary method is low. SUMMARY

[0004] The purpose of the present application is to provide a consultation record summarization method and device based on a large language model (LLM), which can improve the accuracy of the consultation record summarization method.

[0005] To achieve the above-mentioned purpose, the present application provides the following solutions: In a first aspect, the present application provides a consultation record summarization method based on a large language model (LLM), comprising: Collecting multi-source customer service data; wherein the multi-source customer service data at least includes conversation record data, call record data and work order information; Preprocessing the multi-source customer service data to obtain target multi-source customer service data; Using a large language model to process the target multi-source customer service data to obtain initial summary content; Generating standardized consultation record summary content based on the initial summary content and a preset standardized template.

[0006] Optionally, the preprocessing of the multi-source customer service data to obtain target multi-source customer service data specifically includes: Cleaning invalid data in the conversation record data, the call record data and the work order information to obtain initial conversation record data, initial call record data and initial work order information; Converting the format of the initial conversation record data to obtain target conversation record data in a standard format; Performing speech recognition on the initial call record data to obtain target call record data in a standard format; Perform semantic recognition on the initial work order information to obtain target work order information in a standard format; Determine the target conversation record data, the target call record data, and the target work order information as target multi-source customer service data.

[0007] Optionally, the target multi-source customer service data is processed using a large language model to obtain initial summary content, specifically including: Obtain a prompt word template constructed in advance; Input the prompt word template and the target multi-source customer service data into a large language model to obtain initial summary content output by the large language model; wherein the initial summary content at least includes context understanding content, key information, and reasoning content.

[0008] Optionally, after the standardized consultation record summary content is generated, the method further includes: Obtain error content and correction content in the standardized consultation record summary content input by a management personnel; Obtain a to-be-modified prompt word matching the error content from the prompt word template; Determine a target prompt word corresponding to the correction content; Replace the to-be-modified prompt word in the prompt word template with the target prompt word to obtain an updated prompt word template.

[0009] Optionally, the standardized template includes custom field information, custom classification information, and custom summary information, wherein: The custom field information at least includes a field name, a field type, and a field description; The custom classification information at least includes a classification name, a classification condition, and a classification description; The custom summary information at least includes progress stage information, progress description, and completion standard.

[0010] Optionally, the standardized consultation record summary content is generated based on the initial summary content and a preset standardized template, specifically including: Obtain standardized field content matching the custom field information from the initial summary content; Obtain a target classification name matching the initial summary content from the custom classification information; wherein the target classification name corresponds to classification conditions and classification descriptions that match the initial summary content; Obtain a target progress stage matching the initial summary content from the custom summary information; wherein the target progress stage corresponds to progress description and completion standard that match the initial summary content; generate a standardized consultation record summary content including the standardized field content, the target classification name, and the target progress stage.

[0011] In a second aspect, the present application provides a consultation record summary device based on an LLM, comprising: a collection unit configured to collect multi-source customer service data, wherein the multi-source customer service data at least includes conversation record data, call record data, and work order information; a preprocessing unit configured to preprocess the multi-source customer service data to obtain target multi-source customer service data; a processing unit configured to process the target multi-source customer service data using a large language model to obtain initial summary content; a generation unit configured to generate a standardized consultation record summary content based on the initial summary content and a preset standardized template.

[0012] In a third aspect, the present application provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the consultation record summary method based on an LLM according to any one of the above embodiments.

[0013] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program is executable by a processor to implement the steps of the consultation record summary method based on an LLM according to any one of the above embodiments.

[0014] In a fifth aspect, the present application provides a computer program product comprising a computer program, wherein the computer program is executable by a processor to implement the steps of the consultation record summary method based on an LLM according to any one of the above embodiments.

[0015] In a sixth aspect, the present application provides a chip, comprising a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is configured to run a program or instructions, and the processor executes the program or instructions to implement the steps of the consultation record summary method based on an LLM according to any one of the above embodiments.

[0016] According to the embodiments of the present application, the following technical effects are disclosed: The application provides an LLM-based consultation record summarization method, device, equipment, medium and product. By collecting multi-source customer service data, covering conversations, call records and work order information, etc., various key information of customer consultation can be comprehensively obtained, and one-sidedness caused by a single data source can be avoided. The multi-source data is preprocessed to remove noise and unify the format, providing a high-quality data basis for subsequent processing. The powerful language understanding and generation capability of the large language model is used to process the target data, which can accurately extract core points and generate initial summary content that fits the actual situation. Finally, the final summary is generated based on the preset standardized template to ensure that the content meets the specifications and is highlighted. The whole process from data acquisition to final summary generation is closely coordinated and optimized at each stage, effectively reducing the negligence and bias that may occur in manual summarization, thereby greatly improving the accuracy of consultation record summarization. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0018] Figure 1 A flowchart of an LLM-based consultation record summarization method provided by an embodiment of the present application is shown in the figure. Figure 2 A functional module diagram of an LLM-based consultation record summarization device provided by an embodiment of the present application is shown in the figure. Figure 3 A structural diagram of a computer device provided by an embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0020] The above-mentioned purposes, features and advantages of the present application can be more obvious and easy to understand. The present application will be further described in detail below with reference to the drawings and specific embodiments.

[0021] In an exemplary embodiment, as Figure 1As shown, a LLM-based consultation record summarization method is provided, which is executed by a computer device, specifically, can be executed by a terminal or a server, or both, in the embodiments of the present application, the method includes the following steps 101-104. Wherein: Step 101, collecting multi-source customer service data.

[0022] In the embodiments of the present application, the multi-source customer service data includes at least conversation record data, call record data and work order information; wherein: The conversation record data can include text chat records, emoticons, file transfer records; The call record data can include voice-to-text content, call duration, call quality information; The work order information can include work order type, processing status, associated product information, customer level information.

[0023] Step 102, pre-processing the multi-source customer service data to obtain target multi-source customer service data.

[0024] In the embodiments of the present application, the conversation record data, call record data and work order information have inconsistent structures, and the source data contains more respective business attribute fields, which are meaningless for the large model to generate summary records. It is necessary to clean up invalid data and only keep the core records. After data format conversion, the standard format is formed and is recognized and processed by the large model. The standard format can be text format, video format, table format, etc. In this regard, the embodiments of the present application are not limited. Because the text format has the characteristics of small memory space occupation and convenient processing, the text format is usually used as the standard format, and the text used in the text format can be Chinese, English and other types of text, which is not limited in the embodiments of the present application.

[0025] As an optional implementation, the way of step 102 for pre-processing the multi-source customer service data to obtain target multi-source customer service data can include: Cleaning up invalid data in the conversation record data, the call record data and the work order information to obtain initial conversation record data, initial call record data and initial work order information; Converting the format of the initial conversation record data to obtain target conversation record data in a standard format; Performing voice recognition on the initial call record data to obtain target call record data in a standard format; Performing semantic recognition on the initial work order information to obtain target work order information in a standard format; Determining the target conversation record data, the target call record data and the target work order information as the target multi-source customer service data.

[0026] By implementing this embodiment, invalid data is cleaned up, interference information is removed, and subsequent processing is focused on valid content to avoid summary deviation caused by invalid data. Format conversion, speech recognition, and semantic recognition respectively unify the conversation, call, and work order information into a standard format, enhancing the consistency and standardization of the data and facilitating processing by a large language model. The resulting target multi-source customer service data is of high quality and is standardized, laying a solid foundation for subsequent generation of accurate and comprehensive consultation record summaries and effectively ensuring the reliability and practicality of the summary results.

[0027] In step 103, a large language model is used to process the target multi-source customer service data to obtain initial summary content.

[0028] In the embodiments of the present application, a prompt word template specifically for the customer service scenario can be pre-constructed. The large model outputs service summary content, which needs to pay attention to not only the template prompt words set by the customer but also the prompt words of the large model itself. First, the large model is limited to the customer service scenario, and the attributes such as role and emotion are limited. Second, the customer-defined prompt words are avoided from being too broad, and the summary content formed based on the data is avoided from being biased and negative.

[0029] As an optional implementation, the manner in which the large language model is used to process the target multi-source customer service data to obtain initial summary content in step 103 can include: obtaining a pre-constructed prompt word template; inputting the prompt word template and the target multi-source customer service data into a large language model to obtain initial summary content output by the large language model; wherein the initial summary content at least includes context understanding content, key information, and reasoning content.

[0030] By implementing this embodiment, the prompt word template is pre-constructed, which can provide clear and accurate processing guidance for the large language model and make it clear about the task direction. The template and the data are input into the model, which can fully utilize the powerful language processing capability of the model and deeply mine the data connotation. The obtained initial summary content covers context understanding, key information, and reasoning content, which is comprehensive and accurate. It not only presents the overall picture of the consultation but also analyzes and reasons in depth, effectively avoids information omission and misunderstanding, provides strong support for subsequent generation of high-quality consultation record summaries, and improves the overall summary effect.

[0031] In the embodiments of the present application, the context understanding content can be based on complete dialogue history to understand the user's intent; The key information includes identified and extracted key information such as consultation category, problem description, and processing result; The reasoning content is based on the dialogue content to reason the user's real demand and satisfaction.

[0032] The custom field information or the custom classification information in the standardized template has limited the inference target of the large model, which is the customer demand and service satisfaction. For example, the custom classification information may limit pre-sales, after-sales, etc., and the large model only needs to infer the belonging classification according to the data content.

[0033] In step 104, a standardized consultation record summary content is generated based on the initial summary content and a preset standardized template.

[0034] In the embodiments of the present application, the standardized template includes custom field information, custom classification information, and custom summary information, wherein: The custom field information at least includes a field name, a field type, and a field description; The custom classification information at least includes a classification name, a classification condition, and a classification description; The custom summary information at least includes progress stage information, progress description, and completion standard.

[0035] The user can generate a standardized template based on the dynamic prompt words of the custom template. It mainly includes the description information of the template itself, the limit words of the large model, the description information of the template itself is the background information for the large model, which sets the direction of analyzing data; the limit words of the large model are given specific range or boundary to avoid breaking the boundary. It also includes language, emotion, role, etc. Setting, so that the summary given by the large model can better meet the set target.

[0036] As an optional implementation, the manner in which step 104 generates a standardized consultation record summary content based on the initial summary content and a preset standardized template can include: Obtaining standardized field content matched with the custom field information from the initial summary content; Obtaining a target classification name matched with the initial summary content from the custom classification information; wherein the classification condition and the classification description corresponding to the target classification name are both matched with the initial summary content; Obtaining a target progress stage matched with the initial summary content from the custom summary information; wherein the progress description and the completion standard corresponding to the target progress stage are both matched with the initial summary content; Generating a standardized consultation record summary content containing the standardized field content, the target classification name, and the target progress stage.

[0037] Wherein, by implementing this embodiment, the content matching the custom field, classification and summary information, such as standardized field content, target classification name and target progress stage, is accurately obtained from the initial summary content, ensuring that each part of the information closely matches the initial summary and avoiding misplacement of information. By strictly matching the classification conditions and description, progress description and completion standard, the summary content classification is accurate and the progress definition is clear. The final generated standardized summary content is complete, standardized and accurate, which not only meets the preset requirements, but also truly reflects the consulting situation, providing a reliable basis for subsequent query, analysis and utilization.

[0038] In addition, the generated standardized consulting record summary content can also be quality scored; missing information or incorrect information in the standardized consulting record summary content is identified; the LLM model performance is continuously optimized based on artificial feedback; and a summary quality monitoring system is established.

[0039] Wherein, the standardized template, field and other description or prompt word templates can be continuously enriched and optimized by combining actual data and standardized consulting record summary content in an artificial manner, continuously improving the accuracy of LLM summary; Identifying missing or incorrect information, the ultimate goal is to make the summary more accurate by optimizing the prompt words.

[0040] As an optional embodiment, after step 104, the following steps can also be performed: Obtaining the error content and correction content in the standardized consulting record summary content input by the management personnel; Obtaining the to-be-modified prompt word matching the error content from the prompt word template; Determining the target prompt word corresponding to the correction content; Replacing the to-be-modified prompt word in the prompt word template with the target prompt word to obtain an updated prompt word template.

[0041] Wherein, by implementing this embodiment, the errors and correction content pointed out by the management personnel can accurately locate the deviation of the initial summary. The matching to-be-modified prompt word is found from the prompt word template, and the target prompt word corresponding to the correction content is determined to accurately connect the problem and the solution. Replacing the to-be-modified prompt word with the target prompt word obtains an updated template, which can timely correct the model processing logic and avoid the same error from occurring again. This process forms a feedback loop, continuously improving the accuracy of the summary content generated by the large language model, making the subsequent consulting record summary more in line with actual needs, and improving overall work efficiency and quality.

[0042] Implementing steps 101 to 104 effectively reduces potential oversights and biases that may occur during manual summarization, thereby significantly improving the accuracy of the consultation record summary. Furthermore, this application lays a solid foundation for generating accurate and comprehensive consultation record summaries, strongly guaranteeing the reliability and practicality of the summary results. In addition, this application effectively avoids information omissions and misunderstandings, providing strong support for generating high-quality consultation record summaries and improving the overall summary effect. Moreover, this application makes subsequent consultation record summaries more aligned with actual needs, improving overall work efficiency and quality. Furthermore, this application accurately reflects the consultation situation, providing a reliable basis for subsequent querying, analysis, and utilization.

[0043] Based on the same inventive concept, this application also provides an LLM-based consultation record summarization device for implementing the LLM-based consultation record summarization method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations of one or more LLM-based consultation record summarization device embodiments provided below can be found in the limitations of the LLM-based consultation record summarization method described above, and will not be repeated here.

[0044] In one exemplary embodiment, such as Figure 2 As shown, an LLM-based consultation record summarization device is provided, comprising: The data collection unit 201 is used to collect multi-source customer service data; wherein, the multi-source customer service data includes at least conversation record data, call record data, and work order information; Preprocessing unit 202 is used to preprocess the multi-source customer service data to obtain target multi-source customer service data; Processing unit 203 is used to process the target multi-source customer service data using a large language model to obtain initial summary content; The generation unit 204 is used to generate standardized consultation record summary content based on the initial summary content and the preset standardized template.

[0045] In this embodiment of the application, the standardized template includes custom field information, custom category information, and custom summary information, wherein: The custom field information includes at least the field name, field type, and field description; The custom classification information includes at least the classification name, classification criteria, and classification description; The customized summary information includes at least progress stage information, progress description, and completion criteria.

[0046] As an optional implementation, the preprocessing unit 202 preprocesses the multi-source customer service data to obtain target multi-source customer service data in the following manner: cleaning invalid data in the session record data, the call record data, and the work order information to obtain initial session record data, initial call record data, and initial work order information; performing format conversion on the initial session record data to obtain target session record data in a standard format; performing voice recognition on the initial call record data to obtain target call record data in a standard format; performing semantic recognition on the initial work order information to obtain target work order information in a standard format; determining the target session record data, the target call record data, and the target work order information as target multi-source customer service data.

[0047] In this implementation, invalid data is cleaned to remove interference information and focus subsequent processing on valid content, avoiding summary deviation caused by invalid data. Format conversion, voice recognition, and semantic recognition respectively unify session, call, and work order information into a standard format, enhancing data consistency and standardization and facilitating processing by a large language model. The obtained target multi-source customer service data is of high quality and is standardized, laying a solid foundation for subsequent generation of accurate and comprehensive consultation record summaries and effectively ensuring the reliability and practicality of the summary results.

[0048] As an optional implementation, the processing unit 203 processes the target multi-source customer service data using a large language model to obtain initial summary content in the following manner: obtaining a pre-constructed prompt word template; inputting the prompt word template and the target multi-source customer service data into a large language model to obtain initial summary content output by the large language model; wherein the initial summary content at least includes context understanding content, key information, and reasoning content.

[0049] In this implementation, the prompt word template is pre-constructed to provide clear and accurate processing guidance for the large language model, making it clear about the task direction. Inputting the template and the data into the model can fully utilize the powerful language processing capability of the model and deeply mine the data connotation. The obtained initial summary content covers context understanding, key information, and reasoning content, which is comprehensive and accurate, not only presenting the overall picture of the consultation but also analyzing and reasoning in depth, effectively avoiding information omission and misunderstanding, providing strong support for subsequent generation of high-quality consultation record summaries, and improving the overall summary effect.

[0050] As an optional implementation, the generating unit 204 is further configured to: After the standardized consultation record summary content is generated, error content and correction content in the standardized consultation record summary content input by the management personnel are acquired; A to-be-modified prompt word matching the error content is acquired from the prompt word template; A target prompt word corresponding to the correction content is determined; The to-be-modified prompt word in the prompt word template is replaced with the target prompt word, to obtain an updated prompt word template.

[0051] In this implementation, the error and correction content pointed out by the management personnel can be accurately located to position the deviation of the initial summary. The to-be-modified prompt word matching the error content is found from the prompt word template, and the target prompt word corresponding to the correction content is determined, to accurately connect the problem and the solution. The to-be-modified prompt word is replaced with the target prompt word, to obtain an updated template, which can timely correct the model processing logic and avoid the same error from occurring again. This process forms a feedback loop, which continuously improves the accuracy of the summary content generated by the large language model, so that the subsequent consultation record summary is more in line with the actual demand, and the overall work efficiency and quality are improved.

[0052] As an optional implementation, the generation unit 204 generates the standardized consultation record summary content based on the initial summary content and a preset standardized template. The manner of generating the standardized consultation record summary content can be specifically as follows: Standardized field content matching the self-defined field information is acquired from the initial summary content; A target classification name matching the initial summary content is acquired from the self-defined classification information; wherein the classification condition and the classification description corresponding to the target classification name both match the initial summary content; A target progress stage matching the initial summary content is acquired from the self-defined summary information; wherein the progress description and the completion standard corresponding to the target progress stage both match the initial summary content; The standardized consultation record summary content containing the standardized field content, the target classification name, and the target progress stage is generated.

[0053] In this implementation, the content matching the self-defined field, classification, and summary information, such as the standardized field content, the target classification name, and the target progress stage, is accurately acquired from the initial summary content, which ensures that each part of information closely matches the initial summary and avoids information misplacement. The classification condition and the description, and the progress description and the completion standard are strictly matched, so that the summary content is classified accurately and the progress is clearly defined. The finally generated standardized summary content is complete, normative, and accurate, which meets the preset requirements and truly reflects the consultation situation, providing a reliable basis for subsequent query, analysis, and utilization.

[0054] The above-mentioned embodiments effectively reduce the omissions and deviations that may occur in manual summarization, thereby greatly improving the accuracy of the consultation record summary. In addition, the present application also lays a solid foundation for subsequent generation of accurate and comprehensive consultation record summaries, effectively ensuring the reliability and practicality of the summary results. In addition, the present application can effectively avoid information omission and misunderstanding, providing strong support for subsequent generation of high-quality consultation record summaries and improving the overall summary effect. In addition, the present application can make the subsequent consultation record summary more in line with actual needs, improve the overall work efficiency and quality. In addition, the present application can truly reflect the consultation situation, providing a reliable basis for subsequent queries, analysis and utilization.

[0055] In an exemplary embodiment, a computer device is provided, which can be a server or a terminal, and an internal structure diagram thereof can be as shown in Figure 3 The computer device includes a processor, a memory, an input / output interface (I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store LLM-based consultation record summary data. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals through network connection. The computer program is executed by the processor to implement an LLM-based consultation record summary method.

[0056] Those skilled in the art can understand that Figure 3 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0057] In an exemplary embodiment, a computer device is provided, which includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps in the above-mentioned method embodiments.

[0058] In an exemplary embodiment, a computer readable storage medium is provided, storing a computer program, which is executed by a processor to implement the steps in the above-mentioned method embodiments.

[0059] In an example embodiment, a computer program product is provided, including a computer program which, when executed by a processor, implements the steps of any of the above method embodiments.

[0060] In an example embodiment, a chip is provided, including a processor and a communication interface, the communication interface and the processor are coupled, the processor is configured to run a program or an instruction, implement the steps of any of the above method embodiments, and achieve the same technical effects. To avoid repetition, details are not repeated here.

[0061] It should be understood that the chip mentioned in the embodiments of the present application can also be referred to as a system-level chip, a system chip, a chip system, or a system-on-chip chip, etc.

[0062] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant regulations.

[0063] It can be understood by those skilled in the art that all or part of the processes in the above-mentioned embodiments can be completed by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. Any reference to memory, database or other medium used in the embodiments provided by the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (Read-Only Memory, ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive memory (ReRAM), magnetoresistive memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric memory (Ferroelectric Random Access Memory, FRAM), phase change memory (Phase Change Memory, PCM), graphene memory, etc. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), etc.

[0064] The database involved in each of the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a blockchain, and the like, without being limited thereto. The processor involved in each of the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, and the like, without being limited thereto.

[0065] Any combination of the technical features of the above embodiments can be made. In order to make the description simple, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combination of the technical features does not exist, it should be considered as the scope of the present application.

[0066] The principles and implementation modes of the present application are described by applying specific examples herein, and the above embodiments are only used to help understand the method of the present application and its core idea; meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation mode and application range will be changed. In conclusion, the content of the present application should not be understood as a limitation.

Claims

1. A method for summarizing consultation records based on LLM, characterized in that, The LLM-based consultation record summarization method includes: Collect multi-source customer service data; wherein, the multi-source customer service data includes at least conversation record data, call record data, and work order information; The multi-source customer service data is preprocessed to obtain the target multi-source customer service data; The target multi-source customer service data is processed using a large language model to obtain an initial summary. Based on the initial summary content and the preset standardized template, a standardized consultation record summary content is generated.

2. The method for summarizing consultation records based on LLM according to claim 1, characterized in that, The preprocessing of the multi-source customer service data to obtain the target multi-source customer service data specifically includes: The invalid data in the session record data, the call record data, and the work order information are cleaned up to obtain the initial session record data, the initial call record data, and the initial work order information. The initial session record data is converted to a standard format to obtain the target session record data. The initial call log data is subjected to speech recognition to obtain target call log data in a standard format; Semantic recognition is performed on the initial work order information to obtain target work order information in a standard format; The target session record data, the target call record data, and the target work order information are identified as target multi-source customer service data.

3. The method for summarizing consultation records based on LLM according to claim 1, characterized in that, The process of using a large language model to process the target multi-source customer service data to obtain initial summary content specifically includes: Get pre-built prompt word templates; The prompt word template and the target multi-source customer service data are input into the large language model to obtain the initial summary content output by the large language model; wherein, the initial summary content includes at least contextual understanding content, key information and reasoning content.

4. The method for summarizing consultation records based on LLM according to claim 3, characterized in that, After generating the standardized consultation record summary, the method further includes: Obtain the errors and corrections from the standardized consultation record summary input by the management personnel; Obtain the prompt word to be modified that matches the error content from the prompt word template; Identify the target prompt words corresponding to the modified content; Replace the prompt word to be modified in the prompt word template with the target prompt word to obtain the updated prompt word template.

5. The method for summarizing consultation records based on LLM according to claim 1, characterized in that, The standardized template includes custom field information, custom category information, and custom summary information, wherein: The custom field information includes at least the field name, field type, and field description; The custom classification information includes at least the classification name, classification criteria, and classification description; The customized summary information includes at least progress stage information, progress description, and completion criteria.

6. The method for summarizing consultation records based on LLM according to claim 5, characterized in that, The process of generating standardized consultation record summary content based on the initial summary content and the preset standardized template specifically includes: Obtain standardized field content that matches the custom field information from the initial summary content; Obtain the target category name that matches the initial summary content from the custom category information; wherein the category conditions and category description corresponding to the target category name match the initial summary content. Obtain the target progress stage that matches the initial summary content from the custom summary information; wherein the progress description and completion criteria corresponding to the target progress stage match the initial summary content. Generate a standardized consultation record summary that includes the standardized field content, the target category name, and the target progress stage.

7. A consultation record summarization device based on LLM, characterized in that, The LLM-based consultation record summarization device includes: The data collection unit is used to collect multi-source customer service data; wherein, the multi-source customer service data includes at least conversation record data, call record data, and work order information; The preprocessing unit is used to preprocess the multi-source customer service data to obtain target multi-source customer service data; The processing unit is used to process the target multi-source customer service data using a large language model to obtain initial summary content; The generation unit is used to generate standardized consultation record summary content based on the initial summary content and the preset standardized template.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the steps of the LLM-based consultation record summarization method according to any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the LLM-based consultation record summarization method as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the LLM-based consultation record summarization method as described in any one of claims 1-6.

Images