Product recommendation copywriting generation method and device based on artificial intelligence, equipment and medium

Abstract

The invention relates to the technical field of artificial intelligence, and discloses an artificial intelligence-based product recommendation copywriting generation method, device and equipment and a medium, and the artificial intelligence-based product recommendation copywriting generation method comprises the steps of obtaining product information and customer characteristics; converting the customer features into feature statements by using a large language model; inputting the feature statement and the product information into the large language model; and the big language model generates a recommended copywriting according to the feature statements, the product information and trained historical dialogues. According to the invention, the method does not depend on the understanding degree of a salesman on an insurance product and customer characteristics, utilizes massive data information in a large language model, and outputs a copywriting according to the requirements through a guidance model, thereby not only eliminating the requirements for the insurance level of the salesman, but also generating high-quality customized recommendation reasons in batches at the fastest speed.

Description

Technical Field

[0001] The present invention relates to the fields of artificial intelligence technology, natural language processing technology, and financial technology, and in particular to a method, device, equipment and medium for generating product recommendation copy based on artificial intelligence. Background Art

[0002] With the progress of society, in order to meet the needs of different users, the types of insurance products are gradually increasing. Insurance products contain various terms and services, and the documents are generally long. Therefore, when customers are purchasing products, they need to quickly know the product features to determine whether the product meets their needs. At this time, the salesperson needs to show the product features to the customer in concise language.

[0003] Generally speaking, the insurance product data launched by insurance companies is huge and is updated and iterated very quickly. It is difficult for ordinary salesmen to grasp the knowledge of all insurance products in a timely manner. At the same time, the backgrounds of customers faced by salesmen are also different, which makes it very difficult to "recommend the right product to the right customer."

[0004] Existing reasons for product recommendations are mainly written manually by patented salesmen who read the complete copy of the insurance product based on their own experience. This manual writing process is inefficient and requires professional salesmen to complete the writing. It also requires high abilities of the writers, so the labor cost is also high. Summary of the invention

[0005] The present invention provides an artificial intelligence product recommendation copy generation method, device, computer equipment and medium to solve the technical problems of low efficiency of manual writing of product recommendation reasons and high requirements on the ability of the writer.

[0006] In a first aspect, a method for generating product recommendation copy based on artificial intelligence is provided, comprising:

[0007] Obtain product information and customer characteristics;

[0008] Using a large language model to convert the customer features into feature sentences;

[0009] The feature sentences and the product information are input into the large language model, and the large language model generates recommendation texts according to the feature sentences, the product information and the trained historical conversations.

[0010] In a second aspect, a device for generating product recommendation copy based on artificial intelligence is provided, comprising:

[0011] Acquisition module, used to obtain product information and customer characteristics;

[0012] A conversion module, used for converting the customer features into feature sentences using a large language model;

[0013] A generation module is used to input the feature sentence and the product information into the large language model, so that the large language model generates a recommendation copy based on the feature sentence, the product information and the trained historical conversations.

[0014] In a third aspect, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the above-mentioned method for generating product recommendation texts based on artificial intelligence are implemented.

[0015] In a fourth aspect, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the steps of the above-mentioned method for generating product recommendation copy based on artificial intelligence are implemented.

[0016] In the solution implemented by the above-mentioned artificial intelligence-based product recommendation copy generation method, device, computer equipment and storage medium, product information and customer characteristics can be obtained; the customer characteristics are converted into characteristic sentences using a large language model; the characteristic sentences and the product information are input into the large language model; the large language model generates recommendation copy according to the characteristic sentences, the product information and the trained historical dialogue. In the present invention, for insurance business, insurance product recommendation copy can be generated, and by using the product information and the customer characteristics as input, the large language model can automatically output and generate insurance product recommendation copy, and can display targeted product highlights to different customers according to their characteristics, thereby attracting customers to click on products and greatly increasing the probability of customers buying insurance. Compared with the original salesperson writing product recommendation copy according to his own consciousness, the present invention does not rely on the salesperson's understanding of insurance products and customer characteristics. It uses the massive data information in the large language model and outputs the copy according to the requirements by guiding the model, which not only eliminates the requirements for the salesperson's insurance level, but also can generate high-quality customized recommendation reasons in batches at the fastest speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0018] Figure 1 is a schematic diagram of an application environment of a method for generating product recommendation text based on artificial intelligence in one embodiment of the present invention;

[0019] Figure 2 It is a flowchart of a method for generating product recommendation text based on artificial intelligence in one embodiment of the present invention;

[0020] Figure 3 yes Figure 1 A schematic flow chart of a specific implementation of step S100;

[0021] Figure 4 yes Figure 1 A schematic flow chart of a specific implementation of step S200;

[0022] Figure 5 yes Figure 1 A schematic flow chart of a specific implementation of step S400;

[0023] Figure 6 for Figure 5 A schematic flow chart of a specific implementation of step S420;

[0024] Figure 7 for Figure 6 A schematic flow chart of a specific implementation of step S422;

[0025] Figure 8 for Figure 6 A schematic flow chart of a specific implementation of step S423;

[0026] Fig. 9 It is a structural schematic diagram of a device for generating product recommendation text based on artificial intelligence in one embodiment of the present invention;

[0027] Fig.10 is a schematic diagram of a structure of a computer device in one embodiment of the present invention;

[0028] Fig.11 It is another structural schematic diagram of a computer device in one embodiment of the present invention. DETAILED DESCRIPTION

[0029] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0030] The method for generating product recommendation text based on artificial intelligence provided by the embodiment of the present invention can be applied in the following aspects: Figure 1In an application environment, the client communicates with the server through a network. The server can obtain product information and customer features; use a large language model to convert the customer features into feature sentences; input the feature sentences and the product information into the large language model; the large language model generates recommendation texts based on the feature sentences, the product information and the trained historical conversations. The present invention does not rely on the salesperson's understanding of insurance products and customer features. It uses the massive data information in the large language model and guides the model to output texts as required. It not only eliminates the requirements for the salesperson's insurance level, but also can generate high-quality customized recommendation reasons in batches at the fastest speed. Among them, the client can be but is not limited to various personal computers, laptops, smart phones, tablet computers and portable wearable devices. The server can be implemented with an independent server or a server cluster consisting of multiple servers. The present invention is described in detail below through specific embodiments.

[0031] See also Figure 2 As shown, Figure 2 A flowchart of a method for generating product recommendation text based on artificial intelligence provided by an embodiment of the present invention includes the following steps:

[0032] S100: Obtain product information and customer characteristics.

[0033] The artificial intelligence-based product copy recommendation generation method provided by the present invention can be applied to intelligent generation engines such as product auxiliary recommendation platforms in various application scenarios. The intelligent generation engine is usually implemented through a server, which can receive product information and customer characteristics. For example, in the insurance application field, a salesperson can obtain corresponding insurance product information and customer characteristics from the product auxiliary recommendation platform by inputting the customer's ID card or mobile phone number on the product auxiliary recommendation platform.

[0034] For example, if the customer's ID number is entered into the product-assisted recommendation platform, more information about the customer and some insurance product information suitable for the customer can be obtained from the product-assisted recommendation platform through the ID number.

[0035] This method builds an intelligent generation engine that is deployed on the server side and has powerful data processing and analysis capabilities. When the salesperson enters the customer's ID number or mobile phone number on the product recommendation platform, the server side can respond quickly and use advanced artificial intelligence algorithms to extract key information from massive data, including the customer's age, gender, occupation, health status and other characteristics, as well as key elements such as detailed terms, premiums, and coverage of different insurance products.

[0036] Based on this information, the intelligent generation engine can accurately match customer needs with product features and automatically generate personalized insurance product copy. These copies are not only detailed and accurate, but also vivid and easy to understand, which can effectively improve customers' understanding and acceptance of insurance products. At the same time, due to the reduction of manual intervention, the copy production cycle is greatly shortened and the market response speed is improved.

[0037] Among them, Figure 3 As shown, in step S100, the product information is insurance product information, and the step of obtaining product information and customer characteristics includes the following steps:

[0038] S110: Acquire the product information from a product database, where the product information includes the insured amount, premium and / or coverage liability.

[0039] S120: Using the identification information of the customer as a key value, obtaining the customer characteristics presented by the customer in the user portrait library or the team portrait library according to the key value.

[0040] For steps S110-S120, in the intelligent product copy recommendation generation system of the present invention, the product library is a highly structured and detailed database, which stores detailed information of all insurance products, which is the basis for the system to conduct subsequent analysis and recommendations. In order to efficiently obtain this data, the system has designed a specific query interface or API call mechanism. First, when the system needs to obtain information about one or some insurance products, it will send requests through these query interfaces or APIs. These requests may contain specific parameters, such as product ID, product type or other filtering conditions, in order to accurately locate the required product information. Once the request is received, the product library will respond quickly and return a data packet containing the requested product details.

[0041] This product information usually includes but is not limited to key elements such as the product's insured amount, premium, and guarantee liability. The insured amount refers to the maximum compensation amount agreed in the insurance contract, which reflects the insurance company's ability to bear customer risks; the premium is the fee that the customer needs to pay to the insurance company in exchange for insurance protection; and the guarantee liability describes in detail the scope of risks covered by the insurance product, including which risk events can be compensated, as well as the specific standards and conditions for compensation. After obtaining this product information, the system will integrate it into a structured product information data set. This data set not only contains the basic information of each product, but may also include other related attributes, such as product sales, customer reviews, etc. Such a data set provides a solid foundation for subsequent data analysis and product recommendations. Through in-depth mining and analysis of these data, the system can better understand customer needs and provide them with more accurate and personalized insurance product recommendations.

[0042] Secondly, in the intelligent product copy recommendation generation system of the present invention, the construction and retrieval of customer portraits are key steps to achieve accurate recommendations. In order to obtain the detailed characteristics of the customer, the system needs to retrieve the user portrait library and the team portrait library based on the identification information provided by the customer. These identification information may include ID number, mobile phone number, email address or customer ID, etc., which are important bases for the system to identify the customer's identity and obtain relevant information. When the customer enters his identification information, the system will input this information as a key value into the query interface. The user portrait library stores a wealth of customer profiles, purchase history, behavioral habits, preference settings and other information, while the team portrait library contains information such as the customer's role, position, and responsibilities in the team or organization. Through precise queries, the system can quickly extract detailed features that match the customer's identification information from these two databases.

[0043] For example, by analyzing the customer's purchase history, the system can understand the types of insurance products, coverage range, and premium payments that the customer has purchased in the past; by studying the customer's behavioral habits, the system can discover the customer's focus and preferences when browsing insurance products; and by checking the customer's preference settings, the system can learn the customer's specific requirements for coverage period, premium budget, etc. In addition, the information in the team portrait library also helps the system understand the customer's position and influence in the team or organization, so as to provide insurance product recommendations that are more in line with their roles and responsibilities.

[0044] After obtaining this customer feature information, the system will use advanced data analysis technology to deeply mine and process it. By analyzing customer features, the system can evaluate the customer's insurance needs and risk tolerance, and then recommend suitable insurance products for them. This personalized recommendation based on customer portraits can not only meet the customer's personalized needs, but also improve the accuracy and effectiveness of the recommendation, and enhance customer satisfaction and trust in the system.

[0045] The solution can help insurance companies better understand their customers and provide insurance products that meet their needs. In this way, insurance companies can improve customer satisfaction, enhance customer loyalty, and increase sales conversion rates.

[0046] S200: Convert the customer features into feature sentences using a large language model.

[0047] After obtaining detailed feature information of customers from the user portrait library, the collected customer feature information is sorted to ensure the integrity and consistency of the information. This may include converting the feature information into a unified format for subsequent processing, and constructing the sorted customer feature information into a format suitable for input into a large language model. This usually involves converting the feature information into a natural language description, for example: "The customer is a 35-year-old woman who works in the IT industry and has recently shown interest in home property insurance." Send the constructed input format to the large language model (such as GPT, BERT, etc.), requesting the model to generate feature sentences, and the model will generate a coherent text description based on the input feature information. The large language model generates feature sentences based on the input customer features, which can accurately reflect the customer's background and needs. For example: "The customer is a young mother who is concerned about family safety and is looking for insurance products suitable for her family." The generated feature sentences are output and verified to ensure their accuracy and relevance.

[0048] The validity of the generated content can be confirmed through manual review or comparison with business rules. Through this process, the system can transform the discrete feature information of customers into coherent feature sentences, thereby providing support for subsequent personalized recommendations, customer service and marketing strategies. This method not only improves the efficiency of information processing, but also enhances the personalization and pertinence of customer experience.

[0049] It should be noted that BERT (Bidirectional Encoder Representations from Transformers) is a pre-trained language representation model. Unlike traditional unidirectional language models, BERT uses bidirectional training, which means that the model considers the entire context information when processing each word, including the text before and after the word. BERT is based on the Transformer model, which is an architecture based on the self-attention mechanism that can capture the relationship between words. BERT is pre-trained on a large-scale text corpus and has learned rich language knowledge and patterns. BERT can be applied to various downstream tasks such as text classification, sentiment analysis, named entity recognition, etc. through fine-tuning.

[0050] like Figure 4 As shown, the step of converting the customer features into feature sentences using a large language model includes:

[0051] S210, cleaning and formatting the customer characteristics to obtain pre-processed data;

[0052] S220: The large language model converts the preprocessed data into feature sentences of unstructured text type.

[0053] First, the customer characteristics of the customer are collected from the user portrait library. The customer characteristics cover many aspects of the customer, including but not limited to basic information (such as age, gender, occupation, etc.), purchase history (such as which insurance products have been purchased, purchase frequency, etc.), behavior habits (such as browsing preferences, search keywords, etc.) and preference settings (such as coverage period, premium budget, etc.). These data provide the system with comprehensive and in-depth customer insights, which helps to achieve more accurate product recommendations.

[0054] However, raw data often has various problems, such as duplicate records, erroneous data, missing values, noise, etc., which will affect the accuracy of data analysis and recommendations. Therefore, the system needs to clean the collected customer feature data. The cleaning process includes removing duplicate records to avoid data redundancy; correcting erroneous data to ensure data accuracy; filling missing values ​​to avoid analysis bias caused by incomplete data; eliminating noise to improve data purity; standardizing the format to make the data more standardized and easier to process. Formatting may include classifying and organizing data into structured fields, or converting it into a text form of a specific template. After completing data cleaning and formatting, preprocessed data is obtained.

[0055] These data are ready to be input into the big language model. Call the big language model, such as GPT, BERT or other advanced language models, and input the preprocessed data. The big language model generates unstructured text-like feature sentences based on the preprocessed data. These sentences are coherent and easy-to-understand natural language descriptions that can accurately reflect the characteristics of customers. The feature sentences generated by the model can be used in various application scenarios, such as personalized recommendations, customer service, marketing, etc.

[0056] The generated feature sentences are verified to ensure their accuracy and relevance, and the model is adjusted and optimized based on the feedback to improve the quality of the generated sentences. For example, if the customer characteristics include "age: 35 years old, occupation: IT engineer, interests: travel and reading", the large language model may generate the following feature sentence: "This 35-year-old IT engineer likes to travel and read in his spare time." Through this process, the large language model is able to convert structured customer feature data into more humane and easy-to-understand text descriptions.

[0057] S300: Input the feature sentence and the product information into the large language model, and the large language model generates a recommendation text according to the feature sentence, the product information and the trained historical dialogue.

[0058] Combine customer feature statements and product information into a unified input, which includes the customer's basic information, preferences, behavioral data, as well as product features, prices, coverage, etc. Structure the integrated information into an organized data format, such as a JSON object, XML document, or a list of key-value pairs. Structured data helps the model better parse and understand the input content, converting structured data into a model input format, which is a text string, an encoded vector, or a specific data frame. This ensures that large language models receive clear, structured inputs to generate accurate and personalized recommendation copy.

[0059] In the product copy recommendation generation system of the present invention, it is a crucial step to ensure the preprocessing and formatting of all input data. These input data include customer feature statements, product information, and historical conversations. In order to ensure that the model can accurately understand and process these data, the system needs to perform a series of preprocessing and formatting operations on them.

[0060] For customer feature sentences, the system converts them into a form that the model can understand. This usually involves converting feature sentences into text prompts so that the model can generate corresponding copy based on these prompts. In this process, the system needs to ensure the accuracy and completeness of the feature sentences to avoid ambiguity or omissions.

[0061] Next, the system integrates product information into the text prompt. Product information is one of the key elements in generating recommendation copy, which contains the key selling points and related information of the product. By combining product information with text prompts, the system can ensure that the model can fully consider the characteristics and advantages of the product when generating copy, thereby improving the relevance and attractiveness of the copy. In addition, the system will analyze the trained historical conversations. By extracting useful patterns and expressions, the system can provide valuable references for generating new copy. The patterns and expressions in these historical conversations can help the model better understand the needs and preferences of customers, so as to generate copy that better meets customer expectations.

[0062] After completing the above steps, the system will build a text prompt that contains customer characteristics, product information, and historical conversation elements. This text prompt will serve as the basis for the model to generate recommendation copy, guiding the model to generate corresponding copy based on the input information. Finally, the system inputs this text prompt into the big language model. The big language model is a pre-trained model, such as GPT, BERT or other advanced language generation models. These models have powerful natural language processing capabilities and can generate high-quality recommendation copy based on the input text prompt. By leveraging the capabilities of these models, the system can quickly and accurately generate insurance product copy that meets customer needs, improving recommendation results and customer satisfaction.

[0063] This process involves multiple steps, including text generation, screening and optimization of candidate copywriting, and optimization of generated copywriting to ensure the quality and relevance of the copywriting. For example, if the customer feature statement is "This 35-year-old IT engineer is interested in home property insurance", and the product information is "Product A, with a coverage of 1 million, an annual premium of 5,000 yuan, covering home property losses", historical conversations may contain similar customer success stories, then the recommended copywriting generated by the model may be: "Dear IT engineer, we understand that you have special needs for the protection of home property. We recommend that you consider Product A, which provides up to 1 million in coverage, with an annual premium of only 5,000 yuan, and fully covers home property losses. Perfectly aligned with your professional background and family needs, Product A is your ideal choice." Through this process, the relevance and personalization of the recommended copywriting can be ensured, improving customer response rates and satisfaction.

[0064] like Figure 5 As shown, the step of inputting the feature sentence and the product information into the large language model, and the large language model generating a recommendation copy according to the feature sentence, the product information and the trained historical dialogue includes:

[0065] S310, inputting the feature sentences and product information into a large language model;

[0066] S320, guiding the large language model to generate insurance product recommendation copy through instructions and context learning functions; the instructions are constructed by simulated dialogues trained by the large language model, and the simulated dialogues include opening remarks and simulated answers.

[0067] Integrate the preprocessed customer feature statements and product information into an input data set, which will serve as the input of the model. Design an operation instruction, which will guide the model to generate recommended copy. The operation instruction usually includes a description of the task that the large language model needs to perform and any specific format requirements, and construct a simulated dialogue, including an opening statement and simulated answers. The opening statement clarifies the role and task of the model, and the simulated answer shows the style and content of the copy that the model is expected to generate. Format the feature statements, product information, and instructions into an input format that the model can understand, such as JSON or a text string in a specific format. Use an API or directly interact with the model interface to send the formatted input data set to the large language model. Use the contextual learning function of the large language model to ensure that the generated copy is consistent with the provided customer features and product information, and conforms to the context of the insurance industry. Optimize the generated copy to ensure its accuracy, attractiveness, and compliance.

[0068] For example, if the feature sentence is "The customer is interested in home property insurance" and the product information is "Product A provides up to 1 million yuan in coverage with an annual premium of 5,000 yuan", then the simulated dialogue constructed may be:

[0069] Opening remarks: Hello, I am your insurance consultant. How may I help you?

[0070] Model answer: Hi! I noticed that you are interested in homeowners insurance. We have a product that is a good fit for your needs.

[0071] The model might then generate recommendations like:

[0072] "Dear customer, we understand that you have special needs for the protection of your family property. We especially recommend Product A. It provides a coverage of up to 1 million yuan, with an annual premium of only 5,000 yuan, fully covering the safety of your family property. Product A is your ideal choice, providing the most comprehensive protection for you and your family." Through this process, the relevance and personalization of the recommendation copy can be ensured, thereby improving customer response rate and satisfaction.

[0073] like Figure 6 As shown, the step of guiding the large language model to generate insurance product recommendation copy by using instructions and context learning functions includes:

[0074] S321, constructing an opening statement, determining the role played by the large language model and the copywriting template;

[0075] S322, simulating the large language model to answer according to the copy template to generate a sample copy;

[0076] S323. Construct a context learning part according to the example text, so that the large language model generates a text for recommending insurance products.

[0077] In this embodiment, the opening remarks need to clearly define the role played by the model, such as "You are a professional insurance consultant", and the requirements for copy generation, such as "Please generate personalized insurance product recommendation copy based on the customer characteristics and the product information", and clearly define the information points that need to be included in the copy, such as the customer's age, occupation, interests, and product features, prices, and coverage, etc. According to the role and requirements determined in the opening remarks, simulate the answer of the large language model. This includes generating one or more sample copies, which should meet the recommendation scenarios of insurance products, and creating specific sample copies. The sample copy examples will serve as input for the training of the large language model.

[0078] For example: "As a 35-year-old IT engineer, you may need additional protection for your family property. We recommend Product A, which provides up to 1 million yuan in coverage with an annual premium of only 5,000 yuan." Use the generated sample copy to build the context learning part, which includes taking the sample copy as the input of the large language model, allowing the large language model to learn how to generate similar copy based on these examples, inputting the constructed context learning material into the large language model, training the large language model to recognize and understand the structure and content of insurance product recommendation copy, and using the trained model to generate personalized insurance product recommendation copy based on actual customer characteristics and product information. Through this process, it can be ensured that the generated insurance product recommendation copy not only meets customer needs, but is also attractive and persuasive, thereby improving customer response rate and satisfaction.

[0079] like Figure 7 As shown, wherein the step of simulating the large language model to answer based on the copy template to generate an example copy includes:

[0080] S3221, simulating the large language model to answer according to the copy template, and generating multiple first-edition sample copies to cover the diversity of characteristics of different groups of people;

[0081] S3222. Select the first version sample copy that has the highest degree of overlap with the character and copy template of the opening remarks as the sample copy.

[0082] In order to ensure that the generated copy can meet business needs and attract target customers, the system needs to clarify the specific requirements for copy generation. These requirements include target customer groups, product features, expected copy style and tone, etc. First, the system needs to clarify the target customer group. This may involve age, gender, occupation, interests, etc. By deeply understanding the characteristics and needs of target customers, the system can more accurately target the style and content of the copy.

[0083] Secondly, the system needs to highlight the characteristics of the product. This includes information about the product's functions, advantages, price, etc. By incorporating these characteristics into the copy, the system can better demonstrate the value and appeal of the product. In addition, the system needs to determine the expected copywriting style and tone. This may be determined based on factors such as the product's brand image, the cultural background of the target market, and customer preferences. For example, for an insurance product aimed at young people, the copywriting may need to adopt a more lively and fashionable language style; while for an insurance product aimed at corporate customers, the copywriting may need to be more professional and formal.

[0084] After clarifying the copy template, the system will write one or more templates to guide the large language model to generate copy that meets the requirements. These templates will contain the basic structure and key elements of the copy to ensure that the generated copy can meet business needs and customer expectations. Next, the system will create multiple preliminary sample copies. Each sample copy will be designed for different population characteristics, such as age, occupation, interests, etc., to ensure the diversity and coverage of the copy. Then, the system will evaluate each preliminary sample copy and select the copy with the highest degree of overlap with the role and copy template set in the opening remarks as the final sample copy.

[0085] In order to evaluate the quality of the sample copy, the system will develop a series of evaluation criteria. These criteria may include the relevance, accuracy, attractiveness and brand consistency of the copy. By comparing the performance of different sample copies on these criteria, the system can select the best sample copy.

[0086] Once the sample copy is selected, the system will use it to build contextual learning materials. These materials will contain the key information and structure of the sample copy, which will be used to train the large language model to generate recommended copy. In this way, the system can ensure that the generated copy can maintain a certain consistency and quality.

[0087] In actual applications, the system will generate personalized insurance product recommendation copy based on actual customer characteristics and product information. To ensure the quality of these copy, the system will evaluate them to ensure that they meet business goals and customer expectations. At the same time, the system will also optimize the model based on customer feedback to improve the quality and relevance of the copy.

[0088] For example, if the opening line defines the role as “Professional Insurance Advisor,” the first draft of sample copy generated might include:

[0089] "As a young parent, you may be looking for an insurance product that provides the best protection for your children. We recommend Product B, which is specially designed for children and provides comprehensive medical and accidental coverage."

[0090] “For busy professionals, we recommend Product C, which provides a flexible insurance plan to ensure that you can enjoy comprehensive protection while pursuing your career.”

[0091] Select the copy that best matches the role of "professional insurance consultant" as the sample copy for subsequent model training and copy generation.

[0092] like Figure 8 As shown, the step of constructing a context learning part according to the example copy so that the large language model generates a copy for insurance product recommendations includes:

[0093] S3231, inputting the sample text into the large language model as a historical dialogue to generate an output style and form that conforms to the characteristics of different groups of people;

[0094] S3232. Input the customer characteristics and the product information into the large language model as the current dialogue, and the large language model generates insurance product recommendation copy according to the output style and form that conforms to the characteristics of different groups of people.

[0095] A set of sample texts is selected, which can cover the characteristics of different groups of people and match the roles and text templates in the opening remarks, and the selected sample texts are input into the large language model as historical dialogues. The historical dialogues will provide contextual information for the large language model to help the large language model understand different output styles and forms. The large language model is trained using the historical dialogue data so that the large language model can learn the text styles and forms that match the characteristics of different groups of people. The specific characteristics of the customer (such as age, occupation, interests, etc.) and product information (such as insurance amount, premium, coverage, etc.) are input into the large language model as the current dialogue. The large language model generates personalized insurance product recommendation texts based on the input customer characteristics and product information, combined with the output style and form in the historical dialogue. The generated recommendation texts are optimized and adjusted to ensure the accuracy, attractiveness and compliance of the texts, and the generated recommendation texts are evaluated to ensure that they meet business goals and customer expectations. Based on customer feedback and business goals, the model is continuously iterated and improved to improve the quality and relevance of the texts.

[0096] For example, if the customer profile is "a 30-year-old financial professional interested in investment and risk management" and the product information is "Product D, which provides flexible investment options and risk protection", then LLM may generate the following recommendation copy:

[0097] “Dear financial professional, considering your concerns about investment and risk management, we particularly recommend Product D. It offers a variety of investment options while ensuring the safety of your assets. Product D is an ideal choice for you to achieve your financial goals.” Through this process, it can be ensured that the generated insurance product recommendation copy not only meets customer needs, but is also attractive and persuasive, thereby increasing customer response rates and satisfaction.

[0098] It can be seen that in the above scheme, for insurance business, insurance product recommendation copy can be generated. By using the product information and the customer characteristics as input, the large language model can be used to automatically output the recommended copy of the insurance product, and can display the product highlights to different customers according to their characteristics, thereby attracting customers to click on the product, greatly increasing the probability of customers buying insurance. Compared with the original salesperson writing product recommendation copy according to his own consciousness, the present invention does not rely on the salesperson's understanding of insurance products and customer characteristics. It uses the massive data information in the large language model to output the copy according to the requirements through the guidance model. Compared with the traditional manual writing of recommendation copy, the efficiency of the application is higher. It not only eliminates the requirements for the insurance level of the salesperson, but also can generate high-quality customized recommendation reasons in batches at the fastest speed.

[0099] 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.

[0100] In one embodiment, a device for generating product recommendation text based on artificial intelligence is provided, and the device for generating product recommendation text based on artificial intelligence corresponds to the method for generating product recommendation text based on artificial intelligence in the above embodiment. Fig. 9 As shown, the device for generating product recommendation text based on artificial intelligence includes an acquisition model 101, a conversion module 102, an input module 103, and a generation module 104. The detailed description of each functional module is as follows:

[0101] Acquisition module 101, used to acquire product information and customer characteristics;

[0102] A conversion module 102, configured to convert the customer features into feature sentences using a large language model;

[0103] The generating module 103 is used to input the characteristic sentence and the product information into the large language model, so that the large language model generates a recommendation copy according to the characteristic sentence, the product information and the trained historical dialogue.

[0104] In one embodiment, the acquisition module 101 is specifically configured to:

[0105] Acquire the product information from the product database, wherein the product information includes the insured amount, premium and / or coverage liability;

[0106] The customer's identification information is used as a key value, and the customer characteristics presented by the customer in the user portrait library or the team portrait library are obtained according to the key value.

[0107] In one embodiment, the conversion module 102 is specifically configured to:

[0108] Cleaning and formatting the customer characteristics to obtain preprocessed data;

[0109] The large language model converts the preprocessed data into feature sentences of the unstructured text class.

[0110] In one embodiment, the generating module 103 is further configured to:

[0111] Inputting the feature sentences and product information into a large language model;

[0112] The large language model is guided to generate insurance product recommendation copy through instructions and context learning functions; the instructions are constructed by simulated dialogues trained by the large language model, and the simulated dialogues include opening remarks and simulated answers.

[0113] In one embodiment, the generating module 103 is specifically configured to:

[0114] Construct the opening statement, determine the role played by the large language model and the copywriting template;

[0115] Simulating the large language model to answer according to the copy template to generate a sample copy;

[0116] A context learning part is constructed according to the example text, so that the large language model generates insurance product recommendation text.

[0117] In one embodiment, the generating module 103 is specifically configured to:

[0118] The large language model is simulated to answer based on the copywriting template, and multiple first-edition sample copies are generated to cover the diversity of characteristics of different groups of people;

[0119] The first version of the sample copy that has the highest degree of overlap with the character and copy template of the opening remarks is selected as the sample copy.

[0120] In one embodiment, the generating module 103 is specifically configured to:

[0121] Input the sample text into the large language model as a historical dialogue to generate an output style and form that meets the characteristics of different groups of people;

[0122] The customer characteristics and the product information are input into the large language model as the current dialogue, and the large language model generates insurance product recommendation copy according to the output style and form that conforms to the characteristics of different groups of people.

[0123] The present invention provides a device for generating product recommendation texts based on artificial intelligence. By using the product information and the customer characteristics as inputs and utilizing the large language model, the device can automatically output and generate recommendation texts for insurance products. The device can display product highlights to different customers in a targeted manner according to their characteristics, thereby attracting customers to click on products and greatly increasing the probability of customers purchasing insurance. The device does not rely on the salesperson's understanding of insurance products and customer characteristics, but utilizes the massive data information in the large language model and outputs texts as required by guiding the model. This not only eliminates the requirement for the salesperson's insurance level, but also can generate high-quality customized recommendation reasons in batches at the fastest speed.

[0124] For the specific limitations of the product recommendation copy generation device based on artificial intelligence, please refer to the limitations of the product copy recommendation generation method based on artificial intelligence above, which will not be repeated here. Each module in the above-mentioned product recommendation copy generation device based on artificial intelligence can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0125] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Fig.10 As shown. The computer device includes a processor, a memory, a network interface and a database connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external client through a network connection. When the computer program is executed by the processor, it realizes the functions or steps on the server side of a product recommendation copy generation method based on artificial intelligence.

[0126] In one embodiment, a computer device is provided. The computer device may be a client, and its internal structure diagram may be as follows: Fig.11 As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements the functions or steps on the client side of a method for generating product recommendation copy based on artificial intelligence

[0127] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program:

[0128] Obtain product information and customer characteristics;

[0129] Using a large language model to convert the customer features into feature sentences;

[0130] The feature sentences and the product information are input into the large language model, and the large language model generates recommendation texts according to the feature sentences, the product information and the trained historical conversations.

[0131] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0132] Obtain product information and customer characteristics;

[0133] Using a large language model to convert the customer features into feature sentences;

[0134] The feature sentences and the product information are input into the large language model, and the large language model generates recommendation texts according to the feature sentences, the product information and the trained historical conversations.

[0135] It should be noted that the above functions or steps that can be implemented by the computer-readable storage medium or computer device can refer to the relevant descriptions on the server side and the client side in the aforementioned method embodiment. To avoid repetition, they will not be described one by one here.

[0136] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0137] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0138] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit them. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some of the technical features therein by equivalents. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. It should be noted that if software tools or components other than those of the company appear in the embodiments of the present application, they are only used for example introduction and do not represent actual use.

Claims

1. A method for generating product recommendation copy based on artificial intelligence, characterized in that: include: Obtain product information and customer characteristics; Using a large language model to convert the customer features into feature sentences; The feature sentences and the product information are input into the large language model, and the large language model generates recommendation texts according to the feature sentences, the product information and the trained historical conversations.

2. The method for generating product recommendation copy based on artificial intelligence according to claim 1, characterized in that: The product information is insurance product information, and the step of obtaining product information and customer characteristics includes: Acquire the product information from the product database, wherein the product information includes the insured amount, premium and / or coverage liability; The customer's identification information is used as a key value, and the customer characteristics presented by the customer in the user portrait library or the team portrait library are obtained according to the key value.

3. The method for generating product recommendation copy based on artificial intelligence according to claim 1, characterized in that: The step of converting the customer features into feature sentences using a large language model comprises: Cleaning and formatting the customer characteristics to obtain preprocessed data; The large language model converts the preprocessed data into feature sentences of the unstructured text class.

4. The method for generating product recommendation copy based on artificial intelligence according to claim 1, characterized in that: The step of inputting the feature sentence and the product information into the large language model, wherein the large language model generates a recommendation text according to the feature sentence, the product information and the trained historical dialogue comprises: Inputting the feature sentences and product information into a large language model; The large language model is guided to generate insurance product recommendation copy through instructions and context learning functions; the instructions are constructed by simulated dialogues trained by the large language model, and the simulated dialogues include opening remarks and simulated answers.

5. The method for generating product recommendation copy based on artificial intelligence according to claim 4, characterized in that: The step of guiding the large language model to generate insurance product recommendation copy through instruction and context learning functions includes: Construct the opening statement, determine the role played by the large language model and the copywriting template; Simulating the large language model to answer according to the copy template to generate a sample copy; A context learning part is constructed according to the example text, so that the large language model generates insurance product recommendation text.

6. The method for generating product recommendation copy based on artificial intelligence according to claim 5, characterized in that: The step of simulating the large language model to answer based on the copywriting template and generating a sample copywriting includes: The large language model is simulated to answer based on the copywriting template, and multiple first-edition sample copies are generated to cover the diversity of characteristics of different groups of people; The first version of the sample copy that has the highest degree of overlap with the character and copy template of the opening remarks is selected as the sample copy.

7. The method for generating product recommendation copy based on artificial intelligence according to claim 5, characterized in that: The step of constructing a context learning part according to the example text so that the large language model generates a text recommendation for an insurance product includes: Input the sample text into the large language model as a historical dialogue to generate an output style and form that conforms to the characteristics of different groups of people; The customer characteristics and the product information are input into the large language model as the current dialogue, and the large language model generates insurance product recommendation copy according to the output style and form that conforms to the characteristics of different groups of people.

8. A device for generating product recommendation copy based on artificial intelligence, characterized in that: include: Acquisition module, used to obtain product information and customer characteristics; A conversion module, used for converting the customer features into feature sentences using a large language model; A generation module is used to input the feature sentence and the product information into the large language model, so that the large language model generates a recommendation copy based on the feature sentence, the product information and the trained historical conversations.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method for generating product recommendation text based on artificial intelligence as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method for generating product recommendation text based on artificial intelligence as described in any one of claims 1 to 7 are implemented.

Images