Method and system for automatically generating car purchase follow-up strategy based on large model

Abstract

The invention discloses a method and system for automatically generating a car purchase follow-up strategy based on a large model, and the method comprises the steps: extracting chat data of a user from a network, and carrying out the preprocessing of the chat data, so as to guarantee that the use requirements of the large model are met; comparing the chat data with a plurality of preset car purchase stages by using a large model, and judging which car purchase stage the user is in; based on preset samples of follow-up strategies of different car purchase stages, performing fine adjustment on the large model to enable the large model to generate different follow-up strategies in different car purchase stages; and deploying a large model, and designing an API interface for inputting chat data and returning a car purchase stage and a follow-up strategy. The technical scheme of the invention aims to deeply utilize conventional formatted tags and dialogue type chat data by means of the capability of a large model, and provides personalized follow-up strategies for users at different stages in a car purchase process. The user experience is improved, and the conversion rate is improved, so that more efficient business operation is realized.

Description

Technical Field

[0001] The present invention relates to the field of information processing technology, and in particular to a method and system for automatically generating a car purchase follow-up strategy based on a large model. Background Art

[0002] When facing users, the current car purchasing platforms of automobile dealers will determine whether the other party is a potential car buyer based on the user portrait, and use this to decide the next follow-up sales strategy. This is currently mainly achieved through machine learning models and expert rule models.

[0003] Machine learning models use machine learning and deep learning technologies to mine and analyze user portrait information, predict users' car purchasing intentions, and then formulate follow-up strategies. However, traditional machine learning models, such as decision trees, SVMs, Bayesian, etc., are mainly used to process structured data. They have relatively weak semantic understanding of natural language and often cannot fully understand the context and meaning when processing chat data. Traditional machine learning models require feature engineering, and in chat data processing, a lot of text preprocessing and feature extraction are required, which is labor-intensive and error-prone. In addition, traditional machine learning models also have problems such as lack of context understanding and difficulty in maintaining conversation status.

[0004] Expert rule models rely on experts’ understanding of user portraits and key tags. Operational experts analyze factors such as whether the customer has determined the brand, budget, model, whether the customer has arrived at the store, whether the customer has found the model they like, and the time of purchase to determine which stage of the car-buying process the user is in. Then, based on the determined car-buying stage, experts will provide users with follow-up strategies. However, traditional rule models usually have difficulty handling complex and uncertain environments. Their expressiveness and flexibility are limited to their pre-set rules and can only perform well in expected situations. When the number of business logic and rules increases, it becomes very difficult to maintain and manage rule models. If there are contradictory or conflicting rules in the rule set, the decision-making process may be hindered by inconsistent or erroneous outputs. Due to its preset nature, rule models may fail when encountering new and unhandled situations. Summary of the invention

[0005] The main purpose of the present invention is to provide a method and system for automatically generating a car purchase follow-up strategy based on a big model, aiming to make in-depth use of conventional formatted tags and conversational chat data with the help of the big model's capabilities, and then provide personalized follow-up strategies for users at different stages of the car purchase process through the machine learning capabilities of the big model. It can provide strong data support for the marketing work of the entire chain, help improve user experience, increase conversion rate, and thus achieve more efficient business operations.

[0006] To achieve the above purpose, the present invention proposes a method for automatically generating a car purchase follow-up strategy based on a large model, comprising:

[0007] Extracting user chat data from the network and preprocessing the chat data to ensure that it meets the requirements for using the large model;

[0008] Using a large model, the chat data is compared with a plurality of preset car purchase stages to determine which car purchase stage the user is in;

[0009] Based on the preset samples of follow-up strategies at different stages of car purchase, fine-tune the big model so that it can generate different follow-up strategies at different stages of car purchase.

[0010] Deploy a large model and design an API interface for inputting chat data and returning the car purchase stage and follow-up strategy.

[0011] In one embodiment of the present invention, after the step of extracting the user's chat data from the network and preprocessing the chat data, the method further includes:

[0012] Obtain public automobile-related information and data to create a knowledge base;

[0013] Vectorize the data in the knowledge base and save it locally.

[0014] In one embodiment of the present invention, the step of fine-tuning the large model based on the preset samples of follow-up strategies at different stages of car purchase so that the large model generates different follow-up strategies at different stages of car purchase includes:

[0015] Create a sample library to store samples of follow-up strategies at different stages of car purchase;

[0016] Fine-tune the large model based on Prompt technology;

[0017] Using RAG technology, the big model generates follow-up strategies based on the user's car purchase stage and data in the knowledge base;

[0018] The samples in the sample library are used to compare and optimize the generated follow-up strategies.

[0019] In one embodiment of the present invention, the step of fine-tuning the large model based on the Prompt technology includes:

[0020] Determine the tasks that need to be completed by this large model;

[0021] Create one or more initial prompts for the task to guide the large model to produce the desired output;

[0022] The big model is trained and tuned using initial prompts, knowledge base, and chat data to minimize the gap between the big model output and the actual target.

[0023] In one embodiment of the present invention, after the step of using the initial prompt, the knowledge base and the chat data to train and adjust the big model to minimize the gap between the output content of the big model and the actual target, the following steps are further included:

[0024] Record the output of the large model and compare it with the original training target to evaluate the effect of the adjusted large model.

[0025] In one embodiment of the present invention, the steps of deploying a large model and designing an API interface for inputting chat data and outputting a car purchase stage and a follow-up strategy include:

[0026] Use Docker to create application containers;

[0027] Build the pipeline process and call the large model;

[0028] Design an API interface to input chat data and output the car purchase stage and follow-up strategy.

[0029] In one embodiment of the present invention, after the step of designing an API interface for inputting chat data and outputting the car purchase stage and follow-up strategy, the step further includes:

[0030] Add the output follow-up strategy to the sample library.

[0031] The present invention also provides a car purchase follow-up system, comprising:

[0032] A data processing unit, used for extracting user chat data from the network and preprocessing the chat data to ensure that it meets the requirements for using the large model;

[0033] A stage determination unit, used to compare the chat data with a plurality of preset car purchase stages using a large model to determine which car purchase stage the user is in;

[0034] A model fine-tuning unit is used to fine-tune the large model according to preset samples of follow-up strategies at different stages of car purchase, so that the large model generates different follow-up strategies at different stages of car purchase;

[0035] The model deployment unit is used to deploy large models and design API interfaces for inputting chat data and returning the car purchase stage and follow-up strategies.

[0036] The technical solution of the present invention makes in-depth use of conventional formatted tags and conversational chat data by leveraging the capabilities of the big model. Through in-depth understanding and processing of these data, it is possible to generate follow-up strategies that meet the needs of users at different stages of car purchase during the operation process. For example, the big model can mine the user's preferences, interests, and possible concerns from the formatted tag data and chat data, and then formulate and adjust the follow-up strategy in a targeted manner. This strategy may include customized communication information, recommending appropriate communication time, providing content that meets user needs, etc. In this way, no matter which stage the user is in the car purchase process, the operation team can provide them with targeted follow-up strategies in a scientific way.

[0037] In addition, the present invention also includes in-depth analysis of chat data to better understand the problems and concerns that users may encounter during the car-buying process, as well as their responses to various contact strategies. This requires a large model to have sufficient ability to understand and process unstructured chat data, so as to provide practical and operational strategic recommendations.

[0038] In summary, the present invention aims to provide personalized follow-up strategies for users at different stages of the car-buying process by effectively integrating and utilizing formatted tags and chat data and leveraging the machine learning capabilities of large models. This will provide strong data support for full-link marketing work, help improve user experience, increase conversion rates, and thus achieve more efficient business operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0040] Figure 1 A schematic flow chart of the steps of a method for automatically generating a car purchase follow-up strategy based on a large model according to the present invention;

[0041] Figure 2 A flowchart of the steps to fine-tune the large model and generate a follow-up strategy;

[0042] Figure 3 A schematic diagram of the steps for fine-tuning a large model based on the Prompt technique;

[0043] Figure 4 A flowchart of the steps for deploying a large model.

[0044] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0046] Reference Figure 1 The present invention proposes a method for automatically generating a car purchase follow-up strategy based on a large model, comprising:

[0047] S100, extracting user chat data from the network and preprocessing the chat data to ensure that it meets the requirements for using the large model;

[0048] S200, using the big model to compare the chat data with multiple preset car purchase stages to determine which car purchase stage the user is in;

[0049] S300, based on the preset samples of follow-up strategies at different stages of car purchase, fine-tune the large model so that the large model generates different follow-up strategies at different stages of car purchase;

[0050] S400, deploys a large model and designs an API interface for inputting chat data, returning the car purchase stage and follow-up strategy.

[0051] It can be understood that in step S100, chat data is first legally obtained, such as telemarketing data, private domain operation conversation data, private domain operation group chat data, etc.; then the data is pre-processed, such as verifying the chat structure, ensuring that the conversation between customer service and customers is arranged in sequence, the customer service and customer roles are accurate, deleting special symbols in the chat content, and replacing super long links, location coordinate information, and other group chat reference information to meet the requirements for using large models.

[0052] Then, step S200 is followed by the car purchase status determination. First, the car purchase status stage is divided, which is performed by business experts. For example, in this example, the car purchase journey is divided into 8 stages: brand recognition → budget clarification → car model clarification → store experience → test drive → car model → car purchase time → transaction, which is a progressive relationship. The customer's car purchase status is annotated using a large model. The large model can be selected from models such as GLM, GPT-3 or Minimax. This is because the present invention mainly needs to process Chinese, and these models have strong expression capabilities and are more convenient for service calls.

[0053] Step S300 analyzes examples of follow-up strategies for different car purchase states based on the divided car purchase stages, the user's car purchase state and the label portrait. These examples are some sample strategies written manually. These examples are used to fine-tune the model so that the model can generate different follow-up strategies under different car purchase states. Each customer has a specific follow-up time point and recommended words, and these factors also need to be considered when generating a follow-up strategy.

[0054] Finally, the model is deployed and used. To ensure the stability and high availability of the model, you may need to use deployment strategies such as load balancing and multiple instances.

[0055] In the embodiment of the present invention, after extracting the user's chat data from the network and preprocessing the chat data in step S100, the following steps are also included:

[0056] S110, obtaining public automobile-related information data and creating a knowledge base;

[0057] S120, vectorize the data in the knowledge base and save it locally.

[0058] Understandably, the automotive domain knowledge base is constructed based on public car configuration information, historical operation data, and car-related article data, so that the big model can generate content based on the knowledge in this plug-in knowledge base. After obtaining the text data of the knowledge base, Text Splitter, Text Chunks, Embedding, and Vector Store are processed in sequence to vectorize the knowledge base and save it locally.

[0059] Reference Figure 2 In the embodiment of the present invention, based on the preset samples of follow-up strategies at different stages of car purchase, the large model is fine-tuned so that the large model generates different follow-up strategies at different stages of car purchase. Step S300 includes:

[0060] S310, creating a sample library for storing samples of follow-up strategies at different stages of car purchase;

[0061] S320, based on Prompt technology, fine-tuning large models;

[0062] S330, using RAG technology, enables the big model to generate follow-up strategies based on the user's car purchase stage and data in the knowledge base;

[0063] S340, using the samples in the sample library to compare and optimize the generated follow-up strategy.

[0064] Understandably, the sample library is used to store preset follow-up strategy samples, which are usually manually written by business experts based on user portraits and different car purchase stages.

[0065] Prompt technology is to provide input to the model to guide it to generate specific output text or instructions. It is a text paragraph provided by the user when interacting with the model, which is used to describe the information, answers, text, etc. that the user wants to obtain from the model. The purpose of prompt is to guide the model to produce the desired response in order to better control the generated output.

[0066] RAG (Retrieval-Augmented Generation) technology is a technology that uses information from private or proprietary data sources to assist in text generation. It combines retrieval models and generation models to generate accurate and contextual answers. PAG technology enables large models to generate content based on knowledge in external knowledge bases. This approach can effectively improve the accuracy of the output follow-up strategy for large models and improve the utilization of knowledge for knowledge bases.

[0067] After the follow-up strategy is generated, it will be compared with the sample cases in the sample library for optimization and adjustment.

[0068] Reference Figure 3 In the embodiment of the present invention, based on the Prompt technology, step S320 of fine-tuning the large model includes:

[0069] S321, determining the tasks to be completed by this large model;

[0070] S322, creating one or more initial prompts for the task to guide the large model to generate the required output;

[0071] S323, use the initial prompts, knowledge base and chat data to train and adjust the big model to minimize the gap between the output content of the big model and the actual target.

[0072] Understandably, the first task to be completed by this model is to determine the method of automatically generating a car purchase follow-up strategy based on a large model. To accomplish this task, create one or more initial prompts, which should be formed in a way that guides the model to generate the required output. For example, role prompts: as an automotive industry market analyst, output appropriate follow-up strategies based on the customer's car purchase journey status through the conversation content between the salesperson and the customer; follow-up suggestion prompts: ask the customer what they are dissatisfied with, ask about the car purchase concerns, and highlight the characteristics and advantages of this product; chat focus prompts: for example, if the customer prefers financial installments, you can suggest to the salesperson that the monthly payment should not be too high. If the customer's profession is a teacher, the communication direction tends to be professional knowledge, logic and more listening.

[0073] The final follow-up strategy is then adjusted using the initial prompt, the plug-in’s knowledge base, and user chat data to minimize the gap between the output of the large model and the actual goal.

[0074] In the embodiment of the present invention, after step S323 of training and adjusting the big model using the initial prompt, the knowledge base and the chat data to minimize the gap between the output content of the big model and the actual target, the method further includes:

[0075] S324, record the output content of the large model and compare it with the original training target to evaluate the effect of the adjusted large model.

[0076] Understandably, the original training goal is the standard or goal that the model is expected to achieve during the training process. In this case, these goals include whether to generate meaningful text, accurately answer questions, provide effective follow-up strategies, etc. After the model is optimized or adjusted, the actual output of the model needs to be compared with these goals to determine whether the model's performance has met expectations. There are many ways to compare, such as: accuracy assessment, checking the accuracy and relevance of the model output, such as whether the answer to the question is correct and whether the generated text meets expectations; diversity check, evaluating the diversity of the model output to ensure that the model can generate corresponding strategies under different inputs; language quality assessment, evaluating the quality of the output content in terms of language fluency, logical coherence, etc. If the performance of the model does not meet expectations, then you need to adjust the prompts, or try different prompts for further optimization and adjustment until the model evaluation meets expectations.

[0077] Reference Figure 4 In the embodiment of the present invention, a large model is deployed and an API interface is designed to input chat data and output the car purchase stage and follow-up strategy. Step S400 includes:

[0078] S410, creating application containers using Docker;

[0079] S420, build the pipeline process and call the large model;

[0080] S430, design an API interface to input chat data and output the car purchase stage and follow-up strategy.

[0081] Understandably, Docker is an open source application container engine that can create a lightweight, portable, self-sufficient container for any application. Simply put, Docker is similar to a container. Various goods are managed through the standardization of containers, and there is no impact between containers. In other words, the Docker platform is a software containerization platform, and the large model in this example is the cargo in the container. The large model is packaged into Docker, and then the large model can be easily transported to other machines for operation. It is also very easy to load, copy, and remove, which is very suitable for software elastic architecture.

[0082] Pipeline refers to the steps of model implementation, including data reading, preprocessing, model creation, evaluation and parameter adjustment, etc., which is often used to simplify design and improve efficiency. In this example, through the pipeline-style Pipeline construction method, multiple steps can be encapsulated into a pipeline object, which is convenient for managing and adjusting large models.

[0083] API (Application Programming Interface) is an application programming interface that allows users to develop without having to access the source code or understand the details of the internal working mechanism. In this example, the chat data is input through the API interface to output the car purchase stage and follow-up strategy.

[0084] In the embodiment of the present invention, after designing an API interface for inputting chat data and outputting the car purchase stage and follow-up strategy step S430, the following steps are also included:

[0085] S440, adding the output follow-up strategy to the sample library.

[0086] Understandably, expanding the sample library with the output follow-up strategies as new samples can improve the training and adjustment effects of the large model, so that the model can better generate personalized follow-up strategies for user chat data, etc.

[0087] The technical solution of the present invention makes in-depth use of conventional formatted tags and conversational chat data by leveraging the capabilities of the big model. Through in-depth understanding and processing of these data, it is possible to generate follow-up strategies that meet the needs of users at different stages of car purchase during the operation process. For example, the big model can mine the user's preferences, interests, and possible concerns from the formatted tag data and chat data, and then formulate and adjust the follow-up strategy in a targeted manner. This strategy may include customized communication information, recommending appropriate communication time, providing content that meets user needs, etc. In this way, no matter which stage the user is in the car purchase process, the operation team can provide them with targeted follow-up strategies in a scientific way.

[0088] In addition, the present invention also includes in-depth analysis of chat data to better understand the problems and concerns that users may encounter during the car-buying process, as well as their responses to various contact strategies. This requires a large model to have sufficient ability to understand and process unstructured chat data, so as to provide practical and operational strategic recommendations.

[0089] In summary, the present invention aims to provide personalized follow-up strategies for users at different stages of the car-buying process by effectively integrating and utilizing formatted tags and chat data and leveraging the machine learning capabilities of large models. This will provide strong data support for full-link marketing work, help improve user experience, increase conversion rates, and thus achieve more efficient business operations.

[0090] The present invention also provides a car purchase follow-up system, comprising:

[0091] A data processing unit, used to extract user chat data from the network and pre-process the chat data to ensure that it meets the requirements for using the large model;

[0092] A stage judgment unit is used to compare the chat data with multiple preset car purchase stages using a large model to determine which car purchase stage the user is in;

[0093] A model fine-tuning unit is used to fine-tune the large model according to preset samples of follow-up strategies at different stages of car purchase, so that the large model generates different follow-up strategies at different stages of car purchase;

[0094] The model deployment unit is used to deploy large models and design API interfaces for inputting chat data and returning the car purchase stage and follow-up strategies.

[0095] Various objects such as various messages / information / equipment / network elements / systems / devices / actions / operations / processes / concepts that may appear in this application are named. It can be understood that these specific names do not constitute a limitation on the relevant objects. The names assigned may change with factors such as scenarios, contexts or usage habits. The understanding of the technical meaning of the technical terms in this application should be mainly determined from the functions and technical effects embodied / executed in the technical scheme.

[0096] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0097] In the embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0098] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0099] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0100] It should also be understood that in various embodiments of the present application, the first, second, etc. are only used to indicate that multiple objects are different. For example, the first time window and the second time window are only used to indicate different time windows. They should not have any impact on the time window itself, and the first, second, etc. mentioned above should not impose any limitations on the embodiments of the present application.

[0101] It should also be understood that in the various embodiments of the present application, unless otherwise specified or there is a logical conflict, the terms and / or descriptions between different embodiments are consistent and can be referenced to each other, and the technical features in different embodiments can be combined to form new embodiments according to their internal logical relationships.

[0102] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a computer-readable storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned computer-readable storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0103] The present application also provides a computer program product, which includes instructions. When the instructions are executed, the network switch and the network system perform operations of the network switch and the network system corresponding to the above method.

[0104] The present application also provides a network system, the system comprising:

[0105] one or more memories for storing instructions; and

[0106] One or more processors are used to call and run the instructions from the memory to execute the method as described above.

[0107] The present application also provides a chip system, which includes a processor for implementing the functions involved in the above content, such as generating, receiving, sending, or processing the data and / or information involved in the above method.

[0108] The chip system may be composed of chips, or may include chips and other discrete devices.

[0109] The processor mentioned in any of the above places can be a CPU, a microprocessor, an ASIC, or one or more integrated circuits for executing programs for controlling the above-mentioned feedback information transmission method.

[0110] In a possible design, the chip system also includes a memory, which is used to store necessary program instructions and data. The processor and the memory can be decoupled and respectively set on different devices, connected by wire or wireless means to support the chip system to implement various functions in the above embodiments. Alternatively, the processor and the memory can also be coupled on the same device.

[0111] Optionally, the computer instructions are stored in a memory.

[0112] Optionally, the memory is a storage unit within the chip, such as a register, a cache, etc. The memory can also be a storage unit within the terminal located outside the chip, such as a ROM or other types of static storage devices that can store static information and instructions, RAM, etc.

[0113] It can be understood that the memory in the present application can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories.

[0114] The nonvolatile memory may be a ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), or a flash memory.

[0115] The volatile memory may be a RAM, which is used as an external cache. There are many different types of RAM, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct memory bus RAM.

[0116] The embodiments of this specific implementation method are all preferred embodiments of the present application, and are not intended to limit the protection scope of the present application. Therefore, all equivalent changes made based on the structure, shape, and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for automatically generating a car purchase follow-up strategy based on a large model, characterized in that: include: Extracting user chat data from the network and preprocessing the chat data to ensure that it meets the requirements for using the large model; Using a large model, the chat data is compared with a plurality of preset car purchase stages to determine which car purchase stage the user is in; Based on the preset samples of follow-up strategies at different stages of car purchase, fine-tune the big model so that it can generate different follow-up strategies at different stages of car purchase. Deploy a large model and design an API interface for inputting chat data and returning the car purchase stage and follow-up strategy.

2. The method for automatically generating a car purchase follow-up strategy based on a large model according to claim 1, characterized in that: After the step of extracting the user's chat data from the network and preprocessing the chat data, the method further includes: Obtain public automobile-related information and data to create a knowledge base; Vectorize the data in the knowledge base and save it locally.

3. The method for automatically generating a car purchase follow-up strategy based on a large model according to claim 2, characterized in that: The step of fine-tuning the large model based on the preset samples of follow-up strategies at different stages of car purchase so that the large model generates different follow-up strategies at different stages of car purchase includes: Create a sample library to store samples of follow-up strategies at different stages of car purchase; Fine-tune the large model based on Prompt technology; Using RAG technology, the big model generates follow-up strategies based on the user's car purchase stage and data in the knowledge base; The samples in the sample library are used to compare and optimize the generated follow-up strategies.

4. The method for automatically generating a car purchase follow-up strategy based on a large model according to claim 3 is characterized in that: The steps of fine-tuning the large model based on the Prompt technology include: Determine the tasks that need to be completed by this large model; Create one or more initial prompts for the task to guide the large model to produce the desired output; The big model is trained and tuned using initial prompts, knowledge base, and chat data to minimize the gap between the big model output and the actual target.

5. The method for automatically generating a car purchase follow-up strategy based on a large model according to claim 4 is characterized in that: After the step of training and adjusting the big model using the initial prompt, the knowledge base and the chat data to minimize the gap between the output content of the big model and the actual target, it also includes: Record the output of the large model and compare it with the original training target to evaluate the effect of the adjusted large model.

6. The method for automatically generating a car purchase follow-up strategy based on a large model according to claim 5, characterized in that: The steps of deploying a large model and designing an API interface for inputting chat data and outputting the car purchase stage and follow-up strategy include: Use Docker to create application containers; Build the pipeline process and call the large model; Design an API interface to input chat data and output the car purchase stage and follow-up strategy.

7. The method for automatically generating a car purchase follow-up strategy based on a large model according to claim 6, characterized in that: After the steps of designing an API interface for inputting chat data and outputting the car purchase stage and follow-up strategy, the following steps are also included: Add the output follow-up strategy to the sample library.

8. A car purchase follow-up system, characterized in that: include: A data processing unit, used for extracting user chat data from the network and preprocessing the chat data to ensure that it meets the requirements for using the large model; A stage determination unit, used to compare the chat data with a plurality of preset car purchase stages using a large model to determine which car purchase stage the user is in; A model fine-tuning unit is used to fine-tune the large model according to preset samples of follow-up strategies at different stages of car purchase, so that the large model generates different follow-up strategies at different stages of car purchase; The model deployment unit is used to deploy large models and design API interfaces for inputting chat data and returning the car purchase stage and follow-up strategies.

Images