Question processing method and device based on large language model, equipment and medium

Abstract

The invention belongs to the technical field of artificial intelligence, and relates to a question processing method and device based on a large language model, computer equipment and a storage medium, and the method comprises the following steps: carrying out word segmentation processing on a question text input by a client to obtain segmented words; performing service field identification processing on the question text by using the segmented words to obtain a corresponding target service field; acquiring a specified example corresponding to the target business field, and acquiring a specified answer corresponding to the specified example; generating a first input sequence based on the specified example, the specified answer and the question text; obtaining a target instruction corresponding to the target business field; generating a second input sequence based on the first input sequence and the target instruction; processing the second input sequence based on the target large language model to obtain a prediction answer; and performing reply processing on the customer based on the predicted answer. In addition, the predicted answer may be stored in a blockchain. According to the invention, the processing efficiency and the processing accuracy of question processing are improved, and the processing cost of question processing is reduced.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence development technology, and in particular to a question processing method, apparatus, computer equipment and storage medium based on a large language model. Background Art

[0002] With the rapid development of information technology, intelligent customer service systems have become an important tool for financial companies to improve the quality of customer service. When interacting with intelligent customer service, customers may ask various questions, which may involve different fields, such as different classification questions or generation questions. In order to provide accurate answers, multiple large language models for different tasks in different fields are usually deployed in intelligent customer service systems. This will enhance the targeted services of intelligent customer service, improve user experience, and increase user stickiness.

[0003] Then, the traditional way of processing questions by intelligent customer service requires using different large language models for different types of questions, and requires human intervention to manually select the corresponding example sets to assist the model in generating answers. This method of manual judgment and selection is not only inefficient, but also easily introduces subjective bias, affecting the accuracy of the answer. Summary of the invention

[0004] The purpose of the embodiments of the present application is to propose a question processing method, device, computer equipment and storage medium based on a large language model to solve the problem that the existing intelligent customer service question processing method needs to use different large language models for different types of questions, and requires human intervention to manually select the corresponding example set to assist the model in generating answers. This method of manual discrimination and selection is not only inefficient, but also easily introduces subjective bias, affecting the technical problem of the accuracy of the answer.

[0005] In order to solve the above technical problems, the embodiment of the present application provides a question processing method based on a large language model, which adopts the following technical solution:

[0006] Receive a question text input by a customer, and perform word segmentation processing on the question text to obtain corresponding word segments;

[0007] Based on a preset business domain identification strategy, the question text is processed for business domain identification using the word segmentation to obtain a target business domain corresponding to the question text;

[0008] Acquire a specified example corresponding to the target business field, and acquire a specified answer corresponding to the specified example;

[0009] Generate a corresponding first input sequence based on the specified example, the specified answer and the question text;

[0010] Obtaining target instructions corresponding to the target business field;

[0011] Generate a corresponding second input sequence based on the first input sequence and the target instruction;

[0012] Processing the second input sequence based on a preset target large language model to obtain a corresponding predicted answer;

[0013] A response process is performed to the customer based on the predicted answer.

[0014] Furthermore, the step of performing business domain identification processing on the question text using the word segmentation based on a preset business domain identification strategy to obtain a target business domain corresponding to the question text specifically includes:

[0015] Extract examples corresponding to the target number from the pre-built example sets of various business fields;

[0016] Calculate the word frequency of each of the word segments in the examples corresponding to each of the example sets;

[0017] Based on all the example sets, calculating the inverse document frequency of each of the word segments;

[0018] Calculating the word frequency and the inverse document frequency based on a preset calculation formula to obtain a domain score for each of the business fields;

[0019] Filter out the designated business area with the highest field score from all the business areas mentioned;

[0020] The designated business field is used as the target business field.

[0021] Furthermore, the step of generating a corresponding first input sequence based on the specified example, the specified answer and the question text specifically includes:

[0022] Get the preset delimiter;

[0023] Based on the separator, the designated example and the designated answer are concatenated correspondingly to obtain a first concatenated sequence;

[0024] Adding the question text to a specified position in the first spliced ​​sequence to obtain a corresponding second spliced ​​sequence;

[0025] The second spliced ​​sequence is used as the first input sequence.

[0026] Furthermore, the step of obtaining the target instruction corresponding to the target business field specifically includes:

[0027] Obtaining task objectives corresponding to the target business area;

[0028] Get the preset instruction writing strategy;

[0029] Perform corresponding instruction writing processing based on the specified writing strategy and the task goal to obtain corresponding initial instructions;

[0030] Performing ambiguity verification on the initial instruction;

[0031] If the initial instruction passes the ambiguity verification, the initial instruction is used as a target instruction corresponding to the target business field.

[0032] Furthermore, the step of generating a corresponding second input sequence based on the first input sequence and the target instruction specifically includes:

[0033] Get the preset splicing strategy;

[0034] Based on the splicing strategy, the first input sequence and the target instruction are spliced ​​to obtain a corresponding third splicing sequence;

[0035] The third spliced ​​sequence is used as the second input sequence.

[0036] Furthermore, the step of replying to the customer based on the predicted answer specifically includes:

[0037] Verifying the accuracy of the predicted answers;

[0038] If the predicted answer passes the accuracy verification, the predicted answer is processed for clarity improvement to obtain a corresponding first answer;

[0039] Performing formatting and optimization processing on the first answer to obtain a corresponding second answer;

[0040] A reply process is performed to the customer based on the second answer.

[0041] Furthermore, the step of replying to the customer based on the second answer specifically includes:

[0042] obtaining customer data of said customer;

[0043] Performing data analysis on the customer data to obtain corresponding customer demand information;

[0044] Performing corresponding personalized adjustment processing on the second answer based on the customer demand information to obtain a target answer;

[0045] Obtaining a preset target communication method;

[0046] Based on the target communication mode, the target answer is returned to the client.

[0047] In order to solve the above technical problems, the embodiment of the present application further provides a question processing device based on a large language model, which adopts the following technical solution:

[0048] A receiving module is used to receive a question text input by a customer and perform word segmentation processing on the question text to obtain corresponding word segments;

[0049] An identification module, configured to perform business domain identification processing on the question text using the word segmentation based on a preset business domain identification strategy, and obtain a target business domain corresponding to the question text;

[0050] A first acquisition module, used to acquire a specified example corresponding to the target business field, and acquire a specified answer corresponding to the specified example;

[0051] A first generating module, configured to generate a corresponding first input sequence based on the specified example, the specified answer and the question text;

[0052] A second acquisition module, used to acquire a target instruction corresponding to the target business field;

[0053] A second generating module, configured to generate a corresponding second input sequence based on the first input sequence and the target instruction;

[0054] A processing module, configured to process the second input sequence based on a preset target large language model to obtain a corresponding predicted answer;

[0055] A reply module is used to reply to the customer based on the predicted answer.

[0056] In order to solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the following technical solution:

[0057] Receive a question text input by a customer, and perform word segmentation processing on the question text to obtain corresponding word segments;

[0058] Based on a preset business domain identification strategy, the question text is processed for business domain identification using the word segmentation to obtain a target business domain corresponding to the question text;

[0059] Acquire a specified example corresponding to the target business field, and acquire a specified answer corresponding to the specified example;

[0060] Generate a corresponding first input sequence based on the specified example, the specified answer and the question text;

[0061] Obtaining target instructions corresponding to the target business field;

[0062] Generate a corresponding second input sequence based on the first input sequence and the target instruction;

[0063] Processing the second input sequence based on a preset target large language model to obtain a corresponding predicted answer;

[0064] A response process is performed to the customer based on the predicted answer.

[0065] In order to solve the above technical problems, the embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solution:

[0066] Receive a question text input by a customer, and perform word segmentation processing on the question text to obtain corresponding word segments;

[0067] Based on a preset business domain identification strategy, the question text is processed for business domain identification using the word segmentation to obtain a target business domain corresponding to the question text;

[0068] Acquire a specified example corresponding to the target business field, and acquire a specified answer corresponding to the specified example;

[0069] Generate a corresponding first input sequence based on the specified example, the specified answer and the question text;

[0070] Obtaining target instructions corresponding to the target business field;

[0071] Generate a corresponding second input sequence based on the first input sequence and the target instruction;

[0072] Processing the second input sequence based on a preset target large language model to obtain a corresponding predicted answer;

[0073] A response process is performed to the customer based on the predicted answer.

[0074] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0075] The present application first receives a question text input by a customer, and performs word segmentation processing on the question text to obtain corresponding word segments; then, based on a preset business domain identification strategy, the word segmentation is used to perform business domain identification processing on the question text to obtain a target business domain corresponding to the question text; then, a specified example corresponding to the target business domain is obtained, and a specified answer corresponding to the specified example is obtained; and a corresponding first input sequence is generated based on the specified example, the specified answer, and the question text; subsequently, a target instruction corresponding to the target business domain is obtained, and a corresponding second input sequence is generated based on the first input sequence and the target instruction; the second input sequence is further processed based on a preset target large language model to obtain a corresponding predicted answer; and finally, a reply is processed to the customer based on the predicted answer. After receiving the question text input by the customer, this application can automatically identify the business field to which the question text belongs based on the use of the business field identification strategy, and can automatically call the specified example corresponding to the target business field, and will further obtain the pre-written target instructions corresponding to the target business field, and then use the target large language model to process the input sequence generated based on the question text, the specified example and the target instruction to generate the corresponding predicted answer, so as to better exert the reasoning ability of the target large language model and improve the accuracy of the generated predicted answer. In addition, it can effectively avoid manual identification of fields and selection of examples, thereby improving the processing efficiency of question processing, and there is no need to construct a large language model for each task type, which effectively reduces the processing cost and complexity of question processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0077] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0078] Figure 2 A flowchart of an embodiment of a question processing method based on a large language model according to the present application;

[0079] Figure 3 is a structural schematic diagram of an embodiment of a question processing device based on a large language model according to the present application;

[0080] Figure 4 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0081] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0082] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0083] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0084] like Figure 1 As shown, the system architecture 100 may include a terminal device 101, a network 102 and a server 103. The terminal device 101 may be a laptop 1011, a tablet computer 1012 or a mobile phone 1013. The network 102 is used to provide a medium for a communication link between the terminal device 101 and the server 103. The network 102 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0085] The user can use the terminal device 101 to interact with the server 103 through the network 102 to receive or send messages, etc. Various communication client applications can be installed on the terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0086] The terminal device 101 can be various electronic devices with a display screen and supporting web browsing. In addition to the laptop computer 1011, the tablet computer 1012 or the mobile phone 1013, the terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 (Moving Picture Experts Group Audio Layer IV) player, a laptop computer, a desktop computer, etc.

[0087] The server 103 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal device 101 .

[0088] It should be noted that the question processing method based on the large language model provided in the embodiment of the present application is generally executed by a server / terminal device, and accordingly, the question processing apparatus based on the large language model is generally arranged in the server / terminal device.

[0089] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0090] Continue to refer Figure 2 , showing a flowchart of an embodiment of a question processing method based on a large language model according to the present application. According to different requirements, the order of the steps in the flowchart can be changed, and some steps can be omitted. The question processing method based on a large language model provided in the embodiment of the present application can be applied to any scenario where question processing is required, and the question processing method based on a large language model can be applied to products in these scenarios, for example, question processing in the financial insurance field. The question processing method based on a large language model comprises the following steps:

[0091] Step S201, receiving a question text input by a customer, and performing word segmentation processing on the question text to obtain corresponding word segments.

[0092] In this embodiment, the question processing method based on the large language model is executed on the electronic device (eg Figure 1The server / terminal device shown in the figure) can obtain the question text through a wired connection or a wireless connection. It should be noted that the above-mentioned wireless connection method may include but is not limited to 3G / 4G / 5G connection, Wi-Fi connection, Bluetooth connection, Wi MAX connection, Zigbee connection, UWB (Ultra Wi DeBand) connection, and other wireless connection methods currently known or to be developed in the future. The execution subject of this application may be a question processing system, or simply referred to as a system. The above-mentioned question text is various questions raised by customers in the process of interacting with the question processing system. These questions may involve different fields, such as different classification questions or generation questions. Exemplarily, in the business scenario of intelligent question and answer of financial insurance, the above-mentioned question text may include insurance consulting questions, financial product recommendation questions, and so on. In addition, a general word segmentation tool can be used to perform word segmentation processing on the question text to obtain corresponding word segments. The number of word segments includes multiple. The word segmentation can also be filtered and word form restored to improve the accuracy and standardization of the obtained word segmentation.

[0093] Step S202: Based on a preset business domain identification strategy, the word segmentation is used to perform business domain identification processing on the question text to obtain a target business domain corresponding to the question text.

[0094] In this embodiment, the above-mentioned preset business domain identification strategy is based on the word segmentation to perform business domain identification processing on the question text to obtain the specific implementation process of the target business domain corresponding to the question text. This application will provide further detailed descriptions of this in subsequent specific embodiments and will not be elaborated on here.

[0095] Step S203, obtaining a specified example corresponding to the target business field, and obtaining a specified answer corresponding to the specified example.

[0096] In this embodiment, a target example set corresponding to the target business field can be called, and a certain number of examples can be randomly extracted from the target example set as the above-mentioned designated examples, and answers corresponding to questions / queries in each extracted example can be obtained to obtain designated answers corresponding to the designated examples. There is no specific limitation on the selection of the above-mentioned certain number of values, which can be set according to actual usage requirements, for example, it can be set to be consistent with the above-mentioned target number.

[0097] Step S204: Generate a corresponding first input sequence based on the specified example, the specified answer and the question text.

[0098] In this embodiment, the specific implementation process of generating the corresponding first input sequence based on the specified example, the specified answer and the question text will be further described in detail in subsequent specific embodiments of the present application and will not be elaborated on here.

[0099] Step S205: acquiring a target instruction corresponding to the target business field.

[0100] In this embodiment, the specific implementation process of obtaining the target instruction corresponding to the target business field will be further described in detail in subsequent specific embodiments of the present application, and will not be elaborated on here.

[0101] Step S206: Generate a corresponding second input sequence based on the first input sequence and the target instruction.

[0102] In this embodiment, the specific implementation process of generating the corresponding second input sequence based on the first input sequence and the target instruction will be further described in detail in subsequent specific embodiments of the present application, and will not be elaborated on here.

[0103] Step S207: Process the second input sequence based on a preset target large language model to obtain a corresponding predicted answer.

[0104] In this embodiment, the target large language model can be any large language model (LLM). By inputting the second input sequence into the target large language model, the target large language model will internally perform reasoning processing on the second input sequence and generate a corresponding predicted answer. The predicted answer is generated by the target large language model based on the second input sequence, and conforms to the predefined domain characteristics and task objectives corresponding to the target business domain.

[0105] Step S208: Reply to the customer based on the predicted answer.

[0106] In this embodiment, the specific implementation process of replying to the customer based on the predicted answer will be described in further detail in subsequent specific embodiments of the present application and will not be elaborated on here.

[0107] The present application first receives a question text input by a customer, and performs word segmentation processing on the question text to obtain corresponding word segments; then, based on a preset business domain identification strategy, the word segmentation is used to perform business domain identification processing on the question text to obtain a target business domain corresponding to the question text; then, a specified example corresponding to the target business domain is obtained, and a specified answer corresponding to the specified example is obtained; and a corresponding first input sequence is generated based on the specified example, the specified answer, and the question text; subsequently, a target instruction corresponding to the target business domain is obtained, and a corresponding second input sequence is generated based on the first input sequence and the target instruction; the second input sequence is further processed based on a preset target large language model to obtain a corresponding predicted answer; and finally, a reply is processed to the customer based on the predicted answer. After receiving the question text input by the customer, this application can automatically identify the business field to which the question text belongs based on the use of the business field identification strategy, and can automatically call the specified example corresponding to the target business field, and will further obtain the pre-written target instructions corresponding to the target business field, and then use the target large language model to process the input sequence generated based on the question text, the specified example and the target instruction to generate the corresponding predicted answer, so as to better exert the reasoning ability of the target large language model and improve the accuracy of the generated predicted answer. In addition, it can effectively avoid manual identification of fields and selection of examples, thereby improving the processing efficiency of question processing, and there is no need to construct a large language model for each task type, which effectively reduces the processing cost and complexity of question processing.

[0108] In some optional implementations, step S202 includes the following steps:

[0109] Examples corresponding to the target number are extracted from the pre-built example sets of various business fields.

[0110] In this embodiment, for each different business field, a corresponding example set is pre-constructed for each business field, and each example set contains multiple examples. There is no specific limitation on the selection of the numerical value of the above target number, which can be set according to actual business needs. In addition, examples corresponding to the target number can be extracted from the example set of each business field by random sampling.

[0111] The above business domain identification strategy is a processing strategy based on the TF-IDF (Term Frequency-Inverse Document Frequency) algorithm. The TF-IDF algorithm is a classic information retrieval algorithm that takes into account term frequency (TermFrequency) and inverse document frequency (Inverse Document Frequency). Term frequency indicates the number of times a word appears in a document. Inverse document frequency indicates the prevalence of a word in the entire document collection. Specifically, the TF-IDF score calculation formula is an indicator used to measure the importance of a word in a document collection. The calculation formula is as follows: TF-IDF = TF*IDF. Among them, TF (Term Frequency) indicates the frequency of a word in a document, which can be calculated by the following first formula: TF = (number of times a word appears in a document) / (total number of words in a document). IDF (Inverse Document Frequency) indicates inverse document frequency, which can be calculated by the following second formula: IDF = log ((total number of documents in a document collection) / (number of documents containing a word)). The higher the TF-IDF score, the more important the word is in the document collection. The core idea is that when a word appears frequently in a document (high TF value) and rarely appears in other documents (high IDF value), the word is more important to the document.

[0112] The word frequencies of each of the word segments in the examples corresponding to each of the example sets are calculated.

[0113] In this embodiment, the term frequency of each of the word segments in the examples corresponding to each of the example sets may be calculated based on the first formula corresponding to TF (Term Frequency).

[0114] Based on all the example sets, the inverse document frequency of each of the word segments is calculated.

[0115] In this embodiment, the inverse document frequency of each of the word segments can be calculated based on the second formula corresponding to IDF (Inverse Document Frequency).

[0116] The word frequency and the inverse document frequency are calculated based on a preset calculation formula to obtain a domain score for each of the business fields.

[0117] In this embodiment, the calculation formula is specifically the TF-IDF score calculation formula. According to the TF-IDF score calculation formula, the word frequency and the inverse document frequency can be calculated to obtain the domain score of each of the business fields.

[0118] The designated business area with the highest field score is selected from all the business areas mentioned.

[0119] In this embodiment, the domain scores of all business domains may be numerically compared to select a designated business domain with the highest domain score from all the business domains.

[0120] The designated business field is used as the target business field.

[0121] The present application extracts examples corresponding to the target number from the pre-constructed example sets of each business field; then calculates the word frequency of each of the segmented words in the examples corresponding to each of the example sets; then calculates the inverse document frequency of each of the segmented words based on all of the example sets; subsequently calculates the word frequency and the inverse document frequency based on a preset calculation formula to obtain the field score of each of the business fields; further selects the designated business field with the highest field score from all of the business fields; and finally uses the designated business field as the target business field. The present application extracts examples corresponding to the target number from the pre-constructed example sets of each business field, and then performs corresponding calculation processing on the segmented words, examples, and example sets based on the calculation formula to obtain the field score of each of the business fields. Subsequently, the designated business field with the highest field score is selected from all of the business fields and used as the final target business field, thereby realizing the rapid and accurate completion of the business field recognition processing for the question text, improving the processing efficiency and processing accuracy of the business field recognition processing, and ensuring the data accuracy of the obtained target business field.

[0122] In some optional implementations of this embodiment, step S204 includes the following steps:

[0123] Get the default delimiter.

[0124] In this embodiment, the separator may be a line break character or a special mark.

[0125] Based on the separator, the designated example and the designated answer are concatenated accordingly to obtain a first concatenated sequence.

[0126] In this embodiment, each designated example and the corresponding designated answer are concatenated, and the separator is added between them and between each example, so as to obtain the corresponding first concatenated sequence. By separating the example and the answer with a separator, it is possible to effectively ensure that the entire sequence maintains a clear structure and is converted into a format suitable for large model processing.

[0127] The question text is added to a specified position in the first concatenated sequence to obtain a corresponding second concatenated sequence.

[0128] In this embodiment, the designated position may be the end of the sequence. The question text may be added to the end of the sequence in the first concatenated sequence to obtain the corresponding second concatenated sequence.

[0129] The second spliced ​​sequence is used as the first input sequence.

[0130] The present application obtains a preset delimiter; then based on the delimiter, the specified example and the specified answer are spliced ​​together to obtain a first spliced ​​sequence; then the question text is added to a specified position in the first spliced ​​sequence to obtain a corresponding second spliced ​​sequence; and the second spliced ​​sequence is subsequently used as the first input sequence. The present application obtains a first spliced ​​sequence by splicing the specified example and the specified answer together based on the use of a delimiter, and then the question text is added to a specified position in the first spliced ​​sequence, so that the first input sequence with a format suitable for large model processing can be quickly and accurately generated, ensuring the accuracy and standardization of the obtained first input sequence.

[0131] In some optional implementations, step S205 includes the following steps:

[0132] Obtaining a task objective corresponding to the target business domain.

[0133] In this embodiment, the domain characteristics corresponding to the above-mentioned target business domain can be obtained, such as common task types, types of questions that users may ask, and expected answer formats. Based on the domain characteristics, the specific tasks that the target large language model needs to perform in the target business domain are determined, thereby obtaining the corresponding task objectives. For example, if the target business domain is the insurance consulting field, the corresponding tasks may be to classify and explain the types of insurance, or to interpret insurance policies, etc.

[0134] Get the preset instruction writing strategy.

[0135] In this embodiment, the above instruction writing strategy can be constructed according to actual business needs. Specifically, the policy content of the instruction writing strategy may include: writing a clear and concise instruction Inst r uct ion according to the task objectives. The instruction should clearly indicate the field of the question, the task type, and the expected answer method. For example: "Please note that the following questions belong to the field of insurance consulting. Your task is to read and understand the customer's questions carefully, then classify the types of insurance they ask, and give detailed explanations."

[0136] Based on the specified writing strategy and the task goal, corresponding instruction writing processing is performed to obtain corresponding initial instructions.

[0137] In this embodiment, instruction writing processing can be performed according to the policy content of the above-mentioned designated writing policy to generate initial instructions corresponding to the task objectives. For example, if the target business field is the insurance consulting field, the generated initial instructions may include: "This question belongs to the insurance consulting question, classify and explain the types of insurance asked by the customer", or may also include: "Please note that the following questions belong to the insurance consulting field. Your task is to read and understand the customer's questions carefully, and then classify the types of insurance asked by the customer and give a detailed explanation", etc.

[0138] An ambiguity verification is performed on the initial instruction.

[0139] In this embodiment, the above-mentioned ambiguity verification refers to the verification process of reviewing whether the generated initial instruction is ambiguous.

[0140] If the initial instruction passes the ambiguity verification, the initial instruction is used as a target instruction corresponding to the target business field.

[0141] In this embodiment, if it is detected that the generated initial instruction can clearly convey to the target large language model the specific requirements and expectations of the task, it is determined that the initial instruction passes the ambiguity verification, and then the initial instruction is used as the target instruction corresponding to the target business field.

[0142] The present application obtains the task goal corresponding to the target business field; then obtains the preset instruction writing strategy; then performs the instruction writing process corresponding to the task goal based on the specified writing strategy to obtain the corresponding initial instruction; then performs ambiguity verification on the initial instruction; if the initial instruction passes the ambiguity verification, the initial instruction is used as the target instruction corresponding to the target business field. The present application obtains the task goal corresponding to the target business field, and then performs the instruction writing process corresponding to the task goal based on the preset specified writing strategy to obtain the initial instruction, and then intelligently performs ambiguity verification on the initial instruction. If it is detected that the initial instruction passes the ambiguity verification, the initial instruction is used as the target instruction corresponding to the target business field, so as to realize the intelligent and accurate completion of the generation process of the target instruction corresponding to the target business field, and ensure the accuracy and adaptability of the obtained target instruction.

[0143] In some optional implementations, step S206 includes the following steps:

[0144] Get the preset stitching strategy.

[0145] In this embodiment, the strategy content of the above splicing strategy may include: splicing the written target instruction in front of the first input sequence constructed previously to form a complete input sequence Input. When splicing, ensure that there is an obvious separator between the instruction and the sequence so that the large model can distinguish them. The separator may include a line break character \n.

[0146] The first input sequence and the target instruction are spliced ​​based on the splicing strategy to obtain a corresponding third splicing sequence.

[0147] In this embodiment, the first input sequence and the target instruction may be spliced ​​together according to the strategy content of the splicing strategy to obtain a corresponding third splicing sequence.

[0148] The third spliced ​​sequence is used as the second input sequence.

[0149] The present application obtains a preset splicing strategy; then splices the first input sequence and the target instruction based on the splicing strategy to obtain a corresponding third splicing sequence; and subsequently uses the third splicing sequence as the second input sequence. The present application splices the first input sequence and the target instruction based on the use of the splicing strategy, so as to quickly and accurately generate a second input sequence in a format suitable for large model processing, thereby ensuring the accuracy and standardization of the generated first input sequence.

[0150] In some optional implementations of this embodiment, step S208 includes the following steps:

[0151] The predicted answers are verified for accuracy.

[0152] In this embodiment, the accuracy verification includes fact checking, logical coherence checking, and context consistency checking. Fact checking includes: verifying whether the factual information in the predicted answer is accurate, especially the factual information verification when it involves specific information such as data, dates, names, and place names. Logical coherence checking includes: checking whether the logic in the answer is coherent and whether there are any contradictions between the various parts. Context consistency includes: ensuring that the predicted answer is consistent with the context of the original question or conversation and does not deviate from the topic.

[0153] If the predicted answer passes the accuracy verification, the predicted answer is processed for clarity improvement to obtain a corresponding first answer.

[0154] In this embodiment, the clarity enhancement process includes language simplification, segmentation, and highlighting. Specifically, language simplification includes: using simple and easy-to-understand language to express complex concepts, and avoiding overly professional or obscure terms unless the customer group is familiar with these terms. Segmentation includes: splitting long answers into multiple paragraphs or list items, each part revolving around a topic to improve readability. Highlighting includes: using bold, italics, or different colors to highlight key information in the answer to help customers quickly grasp the key content.

[0155] The first answer is formatted and optimized to obtain a corresponding second answer.

[0156] In this embodiment, the above-mentioned formatting optimization processing may specifically include: if the predicted answer contains multiple items or data points, they can be organized into a list or table format to facilitate quick browsing and comparison by the customer. In addition, numbers and titles can be added to the list items or paragraphs so that the customer can easily track and understand the hierarchy of information. In addition, if applicable, pictures, charts or graphs can be added to assist in explaining complex concepts or data, making the answer more intuitive and easy to understand.

[0157] A reply process is performed to the customer based on the second answer.

[0158] In this embodiment, the specific implementation process of replying to the customer based on the second answer will be further described in detail in subsequent specific embodiments of the present application and will not be elaborated on here.

[0159] This application verifies the accuracy of the predicted answer; if the predicted answer passes the accuracy verification, the predicted answer is processed for clarity improvement to obtain a corresponding first answer; the first answer is then formatted and optimized to obtain a corresponding second answer; and the customer is subsequently replied to based on the second answer. After processing the second input sequence based on the target large language model to obtain the corresponding predicted answer, this application will further verify the accuracy of the predicted answer. If the predicted answer passes the accuracy verification, the predicted answer is intelligently processed for clarity improvement and the predicted answer is processed for clarity improvement to obtain a second answer, effectively improving the accuracy and readability of the predicted answer, thereby better meeting the needs and expectations of the customer, so that the subsequent reply to the customer based on the second answer is conducive to improving the quality and satisfaction of the answer, thereby improving the customer's user experience.

[0160] In some optional implementations of this embodiment, the reply processing to the customer based on the second answer includes the following steps:

[0161] Obtain customer data for the customer.

[0162] In this embodiment, the above customer data may include basic information, preference settings, historical behavior and other data of the customer. Among them, the basic information may include name, age, gender, occupation, etc. The customer's preference information, such as favorite colors, styles, brands, hobbies, etc., may be collected through questionnaire surveys, user portrait construction, etc. In addition, the customer's historical behavior data, such as purchase records, browsing history, interactive feedback, etc., may be analyzed to understand the customer's consumption habits and demand trends.

[0163] The customer data is analyzed to obtain corresponding customer demand information.

[0164] In this embodiment, by analyzing customer data, specific needs and potential needs of customers are identified, such as specific requirements for product functions, performance, price, service, etc., so as to obtain corresponding customer demand information.

[0165] Based on the customer demand information, the second answer is adjusted accordingly to obtain a target answer.

[0166] In this embodiment, the above-mentioned personalized adjustment processing includes content customization processing, interaction optimization processing and emotional resonance processing. Specifically, the content customization processing includes: adjusting the specific content of the predicted answer or service, such as language style, information presentation method, recommended products, etc. according to customer demand information. The interaction optimization processing includes: optimizing the interaction process with the customer to make it more in line with the customer's usage habits and expectations. For example, adjust the page layout according to the customer's browsing habits, or provide personalized navigation suggestions. Emotional resonance processing includes: incorporating emotional elements into the predicted answers or services, and enhancing the customer's sense of identity with the brand or product through emotional resonance.

[0167] Get the preset target communication method.

[0168] In this embodiment, there is no specific limitation on the above-mentioned target communication method, which can be selected according to actual usage requirements, for example, a chat interface, email, text message or other communication methods desired by the customer can be used.

[0169] Based on the target communication mode, the target answer is returned to the client.

[0170] In this embodiment, the target answer can be returned to the client by using the target communication method to complete the reply processing for the question text input by the user.

[0171] This application obtains the customer data of the customer; then performs data analysis on the customer data to obtain corresponding customer demand information; then performs corresponding personalized adjustment processing on the second answer based on the customer demand information to obtain a target answer; subsequently obtains a preset target communication method, and based on the target communication method, returns the target answer to the customer. This application obtains customer demand information by performing data analysis on the obtained customer data, and then performs corresponding personalized adjustment processing on the second answer based on the customer demand information to obtain a target answer, so that the generated target answer is closer to the customer's actual situation and preferences, and improves the adaptability of the generated target answer. Subsequently, the target answer is returned to the customer using the target communication method, which not only helps to improve customer experience and satisfaction, but also enhances the connection and interaction with customers.

[0172] In some optional implementations, the customer data obtained is subject to customer consent and complies with relevant laws and policies.

[0173] In addition, the question processing method proposed in this application has a strong transfer ability. It can be combined with the chain-of-thought method to further exert the reasoning ability of the large language model. This synergy enables the model to transfer and organize information between different tasks, thereby achieving better performance in various complex situations. Whether facing tasks that require text generation or mathematical problems that require logical reasoning, the question processing method proposed in this application can provide significant improvements in the reasoning ability of the large language model.

[0174] In addition, the question processing method proposed in this application does not require domain fine-tuning of the large language model, which greatly saves time and economic costs. Traditionally, domain fine-tuning requires a lot of resources and time, but the question processing method proposed in this application no longer requires these tedious fine-tuning steps through in-context lending (I CL) enhanced by the retrieval domain. This means that LLMs can be applied to various tasks more quickly, improving the adaptability of the model while reducing implementation costs, making the above-mentioned question processing method more attractive and feasible.

[0175] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0176] It should be emphasized that in order to further ensure the privacy and security of the above prediction answers, the above prediction answers can also be stored in a node of a blockchain.

[0177] The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0178] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0179] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0180] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0181] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0182] Further references Figure 3 , as a response to the above Figure 2 The present application provides an embodiment of a question processing device based on a large language model, and the device embodiment is similar to Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0183] like Figure 3 As shown, the question processing device 300 based on the large language model described in this embodiment includes: a receiving module 301, a recognition module 302, a first acquisition module 303, a first generation module 304, a second acquisition module 305, a second generation module 306, a processing module 307 and a reply module 308. Among them:

[0184] The receiving module 301 is used to receive the question text input by the customer and perform word segmentation processing on the question text to obtain corresponding word segments;

[0185] An identification module 302 is used to perform business domain identification processing on the question text using the word segmentation based on a preset business domain identification strategy to obtain a target business domain corresponding to the question text;

[0186] A first acquisition module 303 is used to acquire a specified example corresponding to the target business field, and acquire a specified answer corresponding to the specified example;

[0187] A first generating module 304, configured to generate a corresponding first input sequence based on the specified example, the specified answer and the question text;

[0188] A second acquisition module 305 is used to acquire a target instruction corresponding to the target business field;

[0189] A second generating module 306, configured to generate a corresponding second input sequence based on the first input sequence and the target instruction;

[0190] A processing module 307, configured to process the second input sequence based on a preset target large language model to obtain a corresponding predicted answer;

[0191] The reply module 308 is used to reply to the customer based on the predicted answer.

[0192] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the question processing method based on the large language model in the aforementioned implementation mode, and are not described in detail here.

[0193] In some optional implementations of this embodiment, the identification module 302 includes:

[0194] An extraction submodule is used to extract examples corresponding to the target number from the pre-built example sets of various business fields;

[0195] A first calculation submodule, used for calculating the word frequency of each of the segmented words in the examples corresponding to each of the example sets;

[0196] A second calculation submodule, used for calculating the inverse document frequency of each of the word segments based on all the example sets;

[0197] A third calculation submodule, configured to calculate the word frequency and the inverse document frequency based on a preset calculation formula to obtain a domain score for each of the business fields;

[0198] A screening submodule, used for screening out a designated business field with the highest field score from all the business fields;

[0199] The first determination submodule is configured to take the designated business field as the target business field.

[0200] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the question processing method based on the large language model in the aforementioned implementation mode, and are not described in detail here.

[0201] In some optional implementations of this embodiment, the first generating module 304 includes:

[0202] The first acquisition submodule is used to acquire a preset delimiter;

[0203] A first splicing submodule, configured to perform corresponding splicing processing on the specified example and the specified answer based on the separator to obtain a first splicing sequence;

[0204] An adding submodule, used for adding the question text to a specified position in the first splicing sequence to obtain a corresponding second splicing sequence;

[0205] The second determining submodule is configured to use the second spliced ​​sequence as the first input sequence.

[0206] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the question processing method based on the large language model in the aforementioned implementation mode, and are not described in detail here.

[0207] In some optional implementations of this embodiment, the second acquisition module 305 includes:

[0208] A second acquisition submodule is used to acquire the task objectives corresponding to the target business field;

[0209] The third acquisition submodule is used to acquire a preset instruction writing strategy;

[0210] A writing submodule, used to perform corresponding instruction writing processing based on the specified writing strategy and the task target to obtain corresponding initial instructions;

[0211] A first verification submodule, configured to perform ambiguity verification on the initial instruction;

[0212] The third determination submodule is configured to use the initial instruction as a target instruction corresponding to the target business field if the initial instruction passes the ambiguity verification.

[0213] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the question processing method based on the large language model in the aforementioned implementation mode, and are not described in detail here.

[0214] In some optional implementations of this embodiment, the second generating module 306 includes:

[0215] The fourth acquisition submodule is used to obtain a preset splicing strategy;

[0216] A second splicing submodule, configured to splice the first input sequence and the target instruction based on the splicing strategy to obtain a corresponding third splicing sequence;

[0217] The fourth determining submodule is configured to use the third spliced ​​sequence as the second input sequence.

[0218] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the question processing method based on the large language model in the aforementioned implementation mode, and are not described in detail here.

[0219] In some optional implementations of this embodiment, the reply module 308 includes:

[0220] A second verification submodule, used to verify the accuracy of the predicted answer;

[0221] A first processing submodule, configured to perform clarity improvement processing on the predicted answer to obtain a corresponding first answer if the predicted answer passes the accuracy verification;

[0222] A second processing submodule is used to perform formatting and optimization processing on the first answer to obtain a corresponding second answer;

[0223] A reply submodule is used to reply to the customer based on the second answer.

[0224] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the question processing method based on the large language model in the aforementioned implementation mode, and are not described in detail here.

[0225] In some optional implementations of this embodiment, the reply submodule includes:

[0226] A first acquisition unit, configured to acquire customer data of the customer;

[0227] An analysis unit, used to analyze the customer data to obtain corresponding customer demand information;

[0228] An adjustment unit, configured to perform corresponding personalized adjustment processing on the second answer based on the customer demand information to obtain a target answer;

[0229] A second acquisition unit, used to acquire a preset target communication mode;

[0230] A returning unit is used to return the target answer to the client based on the target communication mode.

[0231] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the question processing method based on the large language model in the aforementioned implementation mode, and are not described in detail here.

[0232] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0233] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 4 with components 41-43, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital processors (DSPs), embedded devices, etc.

[0234] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0235] The memory 41 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk, a smart memory card (Smart Med ia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the computer device 4. Of course, the memory 41 can also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions of the question processing method based on a large language model, etc. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.

[0236] The processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to run the computer-readable instructions or process data stored in the memory 41, such as computer-readable instructions for running the question processing method based on the large language model.

[0237] The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0238] The present application also provides another embodiment, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the question processing method based on a large language model as described above.

[0239] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0240] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A question processing method based on a large language model, characterized in that: The steps include: Receive a question text input by a customer, and perform word segmentation processing on the question text to obtain corresponding word segments; Based on a preset business domain identification strategy, the question text is processed for business domain identification using the word segmentation to obtain a target business domain corresponding to the question text; Acquire a specified example corresponding to the target business field, and acquire a specified answer corresponding to the specified example; Generate a corresponding first input sequence based on the specified example, the specified answer and the question text; Obtaining target instructions corresponding to the target business field; Generate a corresponding second input sequence based on the first input sequence and the target instruction; Processing the second input sequence based on a preset target large language model to obtain a corresponding predicted answer; A response process is performed to the customer based on the predicted answer.

2. The question processing method based on a large language model according to claim 1, characterized in that: The step of performing business domain identification processing on the question text using the word segmentation based on the preset business domain identification strategy to obtain the target business domain corresponding to the question text specifically includes: Extract examples corresponding to the target number from the pre-built example sets of various business fields; Calculate the word frequency of each of the word segments in the examples corresponding to each of the example sets; Based on all the example sets, calculating the inverse document frequency of each of the word segments; Calculating the word frequency and the inverse document frequency based on a preset calculation formula to obtain a domain score for each of the business fields; Filter out the designated business area with the highest field score from all the business areas mentioned; The designated business field is used as the target business field.

3. The question processing method based on a large language model according to claim 1, characterized in that: The step of generating a corresponding first input sequence based on the specified example, the specified answer and the question text specifically includes: Get the preset delimiter; Based on the separator, the designated example and the designated answer are concatenated correspondingly to obtain a first concatenated sequence; Adding the question text to a specified position in the first spliced ​​sequence to obtain a corresponding second spliced ​​sequence; The second spliced ​​sequence is used as the first input sequence.

4. The question processing method based on a large language model according to claim 1, characterized in that: The step of obtaining the target instruction corresponding to the target business field specifically includes: Obtaining task objectives corresponding to the target business area; Get the preset instruction writing strategy; Perform corresponding instruction writing processing based on the specified writing strategy and the task goal to obtain corresponding initial instructions; Performing ambiguity verification on the initial instruction; If the initial instruction passes the ambiguity verification, the initial instruction is used as a target instruction corresponding to the target business field.

5. The question processing method based on a large language model according to claim 1, characterized in that: The step of generating a corresponding second input sequence based on the first input sequence and the target instruction specifically includes: Get the preset splicing strategy; Based on the splicing strategy, the first input sequence and the target instruction are spliced ​​to obtain a corresponding third splicing sequence; The third spliced ​​sequence is used as the second input sequence.

6. The question processing method based on a large language model according to claim 1, characterized in that: The step of replying to the customer based on the predicted answer specifically includes: Verifying the accuracy of the predicted answers; If the predicted answer passes the accuracy verification, the predicted answer is processed for clarity improvement to obtain a corresponding first answer; Performing formatting and optimization processing on the first answer to obtain a corresponding second answer; A reply process is performed to the customer based on the second answer.

7. The question processing method based on a large language model according to claim 6, characterized in that: The step of replying to the customer based on the second answer specifically includes: obtaining customer data of said customer; Performing data analysis on the customer data to obtain corresponding customer demand information; Performing corresponding personalized adjustment processing on the second answer based on the customer demand information to obtain a target answer; Obtaining a preset target communication method; Based on the target communication mode, the target answer is returned to the client.

8. A question processing device based on a large language model, characterized in that: include: A receiving module is used to receive a question text input by a customer and perform word segmentation processing on the question text to obtain corresponding word segments; An identification module, configured to perform business domain identification processing on the question text using the word segmentation based on a preset business domain identification strategy, and obtain a target business domain corresponding to the question text; A first acquisition module, used to acquire a specified example corresponding to the target business field, and acquire a specified answer corresponding to the specified example; A first generating module, configured to generate a corresponding first input sequence based on the specified example, the specified answer and the question text; A second acquisition module, used to acquire a target instruction corresponding to the target business field; A second generating module, configured to generate a corresponding second input sequence based on the first input sequence and the target instruction; A processing module, configured to process the second input sequence based on a preset target large language model to obtain a corresponding predicted answer; A reply module is used to reply to the customer based on the predicted answer.

9. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the steps of the question processing method based on a large language model as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the steps of the question processing method based on a large language model as described in any one of claims 1 to 7 are implemented.

Images