Product output method and device, computer equipment and storage medium
Through transfer learning and fine-tuning training based on large language models, combined with user portraits and prompt word generation, the personalized problem of insurance product recommendation is solved, accurate business product recommendation services are achieved, and user experience and recommendation accuracy are improved.
Patent Information
- Application Number
- CN202510675447.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-23
- Publication Date
- 2025-08-19
AI Technical Summary
It is difficult for the existing technology to achieve accurate and personalized insurance product recommendation services, especially in the insurance field, where user needs, risk preferences and product terms are ineffective, resulting in poor recommendation results.
A large language model based on pre-training is adopted to build a product output model through transfer learning and fine-tuning training, receive user information and preference data, build user portraits, and generate target prompt words, and generate personalized recommendation results through the product output model.
It has achieved accurate and personalized business product recommendation services, improved the accuracy and user experience of recommendations, and met the diversified needs of the medical and healthcare and financial technology fields.
Smart Images
Figure CN120508710A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a product output method, device, computer equipment, and computer-readable storage medium. Background Art
[0002] Artificial intelligence (AI) technology has made significant progress in recent years, particularly in natural language processing (NLP). Large language models (LLMs), with their powerful language understanding and generation capabilities, have demonstrated outstanding performance in a variety of application scenarios. These models are capable of handling complex language tasks such as text generation, sentiment analysis, and question-answering systems, revolutionizing the field of natural language processing.
[0003] Although large language models have made significant progress in multiple fields, their application in the insurance sector still faces numerous challenges. Insurance product recommendation systems must comprehensively consider a variety of factors, including the user's basic information, insurance needs, risk preferences, and budget. Furthermore, insurance products often have complex terms and coverage, requiring specialized insurance knowledge to accurately understand and interpret. Currently, combining large language models with insurance expertise to achieve accurate and personalized insurance product recommendation services remains a pressing technical challenge in existing technologies. Of course, in the healthcare and fintech sectors, the process of recommending medical insurance products and financial insurance products also faces the challenge of achieving accurate and personalized insurance product recommendation services.
[0004] Based on this, how to provide a product output method, device, computer equipment and computer-readable storage medium that can achieve accurate and personalized business product (such as insurance product) recommendation services is a problem that urgently needs to be solved by technical personnel in this field. Summary of the Invention
[0005] In view of the above-mentioned deficiencies in the prior art, the purpose of the present invention is to provide a product output method, apparatus, computer equipment and computer-readable storage medium, aiming to solve the problem of how to achieve accurate and personalized business product recommendation services.
[0006] In order to achieve the above object, the present invention adopts the following technical solutions:
[0007] In a first aspect, the present invention provides a product output method, comprising:
[0008] Based on the pre-trained large language model, the product output model is constructed through transfer learning and fine-tuning training;
[0009] Receive user information, preference data, and display strategy input by the user terminal, and build a user profile based on the user information and preference data;
[0010] Constructing target prompt words for the product output model based on the user portrait and a preset prompt word generation strategy;
[0011] The target prompt word is used as input, a product output result is generated through the product output model, and the product output result is sent to the user terminal for display according to the display strategy.
[0012] In a second aspect, the present invention provides a product output device, comprising:
[0013] The model building module is used to build a product output model through transfer learning and fine-tuning training based on a pre-trained large language model;
[0014] A portrait building module is used to receive user information, preference data and display strategy input by the user terminal, and build a user portrait based on the user information and preference data;
[0015] A prompt word construction module, configured to construct target prompt words for the product output model based on the user profile and a preset prompt word generation strategy;
[0016] The result generation module is configured to take the target prompt word as input, generate a product output result through the product output model, and send the product output result to the user terminal for display according to the display strategy.
[0017] In a third aspect, the present invention provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the product output method as described above when executing the computer program.
[0018] In a fourth aspect, the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the product output method as described above.
[0019] Compared with the prior art, the present invention provides a product output method, apparatus, computer device and computer-readable storage medium, wherein a product output model is constructed by using a large language model based on pre-training through transfer learning and fine-tuning training; user information, preference data and display strategy input by a user terminal are received, and a user profile is constructed based on the user information and the preference data; target prompt words of the product output model are constructed based on the user profile and a preset prompt word generation strategy; the target prompt words are used as input, a product output result is generated by the product output model, and the product output result is sent to the user terminal for display according to the display strategy; thereby, the present invention can realize accurate and personalized business product recommendation services. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0021] Figure 1 A schematic diagram of an application environment of a product output method provided by an embodiment of the present invention.
[0022] Figure 2 A flowchart of a product output method provided by one embodiment of the present invention.
[0023] Figure 3 A schematic diagram of a program module of a product output device provided by an embodiment of the present invention.
[0024] Figure 4 A schematic diagram of the structure of a computer device provided in one embodiment of the present invention.
[0025] Figure 5 Another structural diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0026] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.
[0027] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.
[0028] It will also be understood that the term "and / or" used in the present description and appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.
[0029] As used in the present specification and the appended claims, the term "if" may be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" may be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.
[0030] In addition, in the description of the present specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.
[0031] References to "one embodiment" or "some embodiments" in the present specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present invention. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in yet other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "including," "comprising," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.
[0032] It should be understood that the order of execution of the steps in the following embodiments does not necessarily mean the order in which they are executed. The order in which each process is executed should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.
[0033] In order to illustrate the technical solution of the present invention, specific embodiments are provided below.
[0034] An embodiment of the present invention provides a product output method that can be applied to Figure 1In the application environment shown, the client and server communicate via a network. The client includes, but is not limited to, PDAs, desktop computers, laptops, ultra-mobile personal computers (UMPCs), netbooks, cloud computing devices, personal digital assistants (PDAs), and other computer devices. The server can be a standalone server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.
[0035] See also Figure 2 An embodiment of the present invention provides a product output method, wherein the method comprises the following steps:
[0036] S100, based on a pre-trained large language model, builds a product output model through transfer learning and fine-tuning training;
[0037] S200: Receive user information, preference data, and display strategy input by a user terminal, and construct a user profile based on the user information and preference data;
[0038] S300: Constructing target prompt words for the product output model based on the user portrait and a preset prompt word generation strategy;
[0039] S400: Taking the target prompt word as input, generating a product output result through the product output model, and sending the product output result to the user terminal for display according to the display strategy.
[0040] In specific implementation, the product (i.e., business product) output method of this embodiment realizes accurate and personalized business product recommendation service by combining the powerful language processing capabilities of the pre-trained large language model and the personalized characteristics of the user portrait. The specific analysis is as follows:
[0041] 1. Application of pre-trained large language models: Through transfer learning and fine-tuning, pre-trained large language models can understand and generate natural language text, providing a strong language processing foundation for product output. This model can handle complex user demand descriptions and business product information, ensuring the accuracy and flexibility of the recommendation process.
[0042] 2. User Profile Construction: Receive detailed user information and preference data and construct a user profile based on it. This user profile comprehensively reflects key information such as the user's needs, risk appetite, and budget, providing an accurate basis for subsequent personalized recommendations. This personalized data processing approach ensures that recommendations are closely aligned with user needs.
[0043] 3. Target prompt word generation: Based on the user profile and a pre-defined prompt word generation strategy, targeted target prompt words are generated. Target prompt words clearly express the user's personalized needs in natural language and serve as input to the product output model. This prompt word generation strategy not only improves the model's input quality but also enhances the relevance and accuracy of recommendation results.
[0044] 4. Generate and display recommendation results: Target prompt words are input into the product output model to generate product output results that closely match user needs. By displaying recommendations according to the user-specified display strategy, users can access recommended information in the most intuitive and convenient way. This personalized display method further enhances user experience and satisfaction.
[0045] This method not only accurately understands user needs but also generates business product recommendations that closely match user profiles, thus achieving truly personalized recommendation services. This approach not only improves the accuracy and relevance of recommendations but also enhances user engagement and trust, providing strong support for precision marketing of business products and optimizing user experience.
[0046] It is understandable that the product output method provided in the embodiments of the present invention can be applied to product output scenarios related to the medical and health field. The following are some specific examples:
[0047] 1. Personalized health insurance recommendations
[0048] Scenario description: A user enters personal information (e.g., age, gender, occupation, income) and preference data (e.g., product type, budget range, risk preference) on a health insurance platform. The platform uses the recommendation method of the present invention, combined with the user profile and target prompt words, to generate personalized health insurance recommendations.
[0049] Specific examples:
[0050] User information: 30 years old, male, software engineer, annual income of 200,000 yuan.
[0051] Preference data: Need a health insurance policy, budget of 5,000 yuan per year, low risk preference.
[0052] Recommendation result: The platform recommended the "Health Worry-Free Plan A" with a premium of 4,800 yuan / year. The coverage includes major diseases and medical expenses, and is suitable for users with low risk preferences.
[0053] 2. Insurance recommendations for patients with chronic diseases
[0054] Scenario description: Users with chronic diseases (such as diabetes and hypertension) enter detailed health status and treatment history on the platform. The platform builds user profiles based on this information and generates insurance product recommendations suitable for patients with chronic diseases.
[0055] Specific examples:
[0056] User information: 45 years old, female, suffering from diabetes, annual income of 150,000 yuan.
[0057] Preference data: I need a health insurance policy that covers the cost of chronic disease treatment, with an annual budget of 8,000 yuan.
[0058] Recommendation result: The platform recommended the "Chronic Disease Protection Plan" with a premium of 7,800 yuan / year. The coverage includes chronic disease treatment expenses and hospitalization medical expenses, and is suitable for patients with chronic diseases.
[0059] It is understandable that the product output method provided in the embodiments of the present invention can also be applied to product output scenarios related to the financial technology field. The following are some specific examples:
[0060] 1. Recommended investment insurance products
[0061] Scenario description: A user enters personal financial information (such as income, assets, and investment preferences) and insurance needs on a fintech platform. The platform uses the recommendation method of the present invention, combined with user profiles and target prompts, to generate investment insurance product recommendations suitable for the user.
[0062] Specific examples:
[0063] User information: 35 years old, male, working in the financial industry, annual income of 300,000 yuan, assets of 1 million yuan.
[0064] Preference data: Need an insurance product with an investment return function, an annual budget of 20,000 yuan, and a medium risk preference.
[0065] Recommendation result: The platform recommended "Investment-Linked Insurance Plan A" with an annual premium of 18,000 yuan and an expected annualized rate of return of 5%, which is suitable for users with investment needs.
[0066] 2. Insurance planning for high net worth clients
[0067] Scenario description: High-net-worth customers enter detailed financial information (such as asset size, investment portfolio, and wealth inheritance needs) and insurance needs on the fintech platform. The platform builds a user profile based on this information and generates insurance plan recommendations suitable for high-net-worth customers.
[0068] Specific examples:
[0069] User information: 50 years old, male, corporate executive, personal assets of 50 million yuan.
[0070] Preference data: Need an insurance product that covers wealth inheritance and high-end medical protection, with an annual budget of 500,000 yuan.
[0071] Recommendation result: The platform recommended the "Exclusive Insurance Package for High Net Worth Clients" with an annual premium of 480,000 yuan. The coverage includes high-end medical insurance, whole life insurance and wealth inheritance insurance, which is suitable for high net worth clients.
[0072] It can be seen from the above examples that the product output method of the present invention can provide accurate and personalized business product recommendation services based on the needs of users in different fields, effectively meeting the diverse needs of users in the fields of medical health and financial technology.
[0073] Furthermore, in one embodiment, the product output method, wherein the product output model is constructed based on the pre-trained large language model through transfer learning and fine-tuning training, specifically comprises the steps of:
[0074] Collect product recommendation data related to the target business products, and clean, label, and extract features from the product recommendation data to obtain a training data set;
[0075] According to a preset training data partitioning strategy, the training data set is divided into a migration data set, a model training set, a model validation set, and a model test set;
[0076] Performing transfer learning on the pre-trained large language model using the migration dataset, and performing fine-tuning training on the large language model after transfer learning using the model training set;
[0077] Evaluating the large language model after fine-tuning training using the model validation set, and optimizing the large language model according to the evaluation results;
[0078] The optimized large language model is tested on the model test set, and when the large language model meets the recommendation performance requirements of the target business product, a product output model is generated.
[0079] During specific implementation, the specific implementation process of the steps in this embodiment is roughly as follows:
[0080] Step 1: Data collection and preprocessing
[0081] 1. Data Collection: Collect product recommendation data related to the target business product (such as insurance products), including user information, business product details, user purchasing behavior, insurance claims records, etc. This data may come from various channels such as insurance company databases, online platforms, and user surveys.
[0082] 2. Data cleaning: Clean the collected data to remove duplicate records, missing values, and outliers. For example, incomplete user records or incorrect insurance amount data may be deleted.
[0083] 3. Data labeling: Label the cleaned data to clearly define the category or label of each data point. For example, labeling a user as "purchased health insurance" or "did not purchase health insurance."
[0084] 4. Feature extraction: Extract useful features from the labeled data, such as the user's age, income, health status, insurance preferences, etc. These features will serve as input for large language model training.
[0085] Result: A high-quality training dataset is obtained, including cleaned, labeled and feature-extracted data.
[0086] Step 2: Data Partitioning
[0087] 1. Partitioning strategy: According to the preset training data partitioning strategy, the training dataset is divided into a migration dataset, a model training set, a model validation set, and a model test set.
[0088] 2. Data distribution: Allocate data to various datasets according to the training data partitioning strategy. Ensure that the data in each dataset is evenly distributed to prevent data bias from affecting model performance.
[0089] Results: Four independent datasets were obtained, namely the migration dataset, model training set, model validation set and model test set.
[0090] Step 3: Model training and optimization
[0091] 1. Transfer Learning: Utilize a transfer dataset to perform transfer learning on a pre-trained large language model. The goal of transfer learning is to adapt the model to the specific language and data characteristics of the insurance domain.
[0092] 2. Fine-tuning: Based on transfer learning, the model is fine-tuned using the model training set. The purpose of fine-tuning is to further optimize the model so that it can accurately recommend insurance products.
[0093] 3. Model Evaluation: Evaluate the fine-tuned model using the model validation set. Evaluation metrics can include accuracy, recall, and F1 score. Optimize the model based on the evaluation results, such as adjusting hyperparameters or structure.
[0094] 4. Model testing: Test the optimized model on the model test set to verify whether the model meets the recommendation performance requirements of the target business product.
[0095] Result: A product output model is generated that meets the recommended performance requirements.
[0096] Through the above process, this embodiment can realize the entire process from data collection to model generation, ensure the efficient construction of the product output model, and provide users with accurate and personalized business product recommendation services.
[0097] Furthermore, in one embodiment, the product output method, wherein the receiving of user information, preference data, and display strategy input by the user terminal, and constructing a user profile based on the user information and the preference data, specifically comprises the steps of:
[0098] Receive user information, preference data, and display strategy input by a user, wherein the user information includes the user's age, gender, occupation, and income; the preference data includes the user's product demand type, risk preference, and budget range; and the display strategy includes the arrangement, display quantity, and display format of the product output results;
[0099] Preprocessing the user information and the preference data, including removing irrelevant characters and unifying text formats;
[0100] Using natural language processing technology to perform word segmentation and feature extraction on the preprocessed user information and preference data to obtain text features and corresponding feature values of the user information and preference data;
[0101] A user profile of the user is constructed based on the text features and the corresponding feature values.
[0102] Furthermore, the product output method, wherein the use of natural language processing technology to perform word segmentation and feature extraction on the pre-processed user information and preference data to obtain text features and corresponding feature values of the user information and preference data, specifically includes the steps of:
[0103] Using natural language processing technology to perform word segmentation on the pre-processed user information and preference data, dividing the text of the user information and preference data into multiple independent vocabulary units, and generating word segmentation results;
[0104] Through keyword extraction and entity recognition, feature extraction is performed on the word segmentation results to obtain text features and corresponding feature values therein;
[0105] Each of the text features is associated with the feature value corresponding to it, so that each of the text features is in one-to-one correspondence with the feature value corresponding to it.
[0106] During specific implementation, the specific implementation process of the steps in this embodiment is roughly as follows:
[0107] Step 1: User data reception and preprocessing
[0108] 1. Receive user input: This process receives user information, preference data, and display strategies from the user. User information includes age, gender, occupation, and income; preference data includes product demand type, risk appetite, and budget range; and the display strategy includes the arrangement of recommended results, the number of recommendations to be displayed, and the format of the recommendations.
[0109] 2. Data preprocessing: Preprocess the user information and preference data entered by the user, including removing irrelevant characters (such as extra spaces and punctuation marks) and unifying the text format (such as converting all text to lowercase).
[0110] Result: The cleaned user information and preference data are obtained and ready for subsequent processing.
[0111] Step 2: Natural Language Processing and Word Segmentation
[0112] 1. Word segmentation: Use natural language processing technology to perform word segmentation on the pre-processed user information and preference data, split the text into multiple independent vocabulary units, and generate word segmentation results.
[0113] 2. Example:
[0114] User input: "I need a low-risk health insurance with a budget of 5,000 yuan per year."
[0115] Word segmentation results: ["I", "need", "a", "low risk", "of", "health insurance", "budget", "every year", "5000 yuan"].
[0116] Result: Obtain user information and preference data after word segmentation, ready for feature extraction.
[0117] Step 3: Feature extraction and keyword identification
[0118] 1. Keyword extraction: Keywords related to insurance needs, such as "health insurance" and "low risk", are extracted from the word segmentation results using keyword extraction technology (such as the TF-IDF algorithm).
[0119] 2. Entity recognition: Use named entity recognition (NER) technology to extract specific numerical and category information from the word segmentation results, such as "budget", "5,000 yuan", "every year", etc.
[0120] 3. Feature correspondence: Associate and map each text feature obtained with its corresponding feature value, so that each text feature corresponds to its corresponding feature value one-to-one, forming a structured feature set.
[0121] Result: The text features of user information and preference data and their corresponding feature values are obtained.
[0122] Step 4: Build a user profile
[0123] Integrate the extracted text features and feature values into a structured data format (such as a dictionary or JSON object) to form a user profile.
[0124] Result: Generate a comprehensive and structured user profile, providing a basis for subsequent business product recommendations.
[0125] Through the above process, this embodiment can realize the complete process from user input to user portrait construction, ensuring that the product output system can provide accurate and personalized business product recommendation services.
[0126] Furthermore, in one embodiment, the product output method, wherein the target prompt words of the product output model are constructed according to the user portrait and a preset prompt word generation strategy, specifically comprises the steps of:
[0127] Analyze the user profile to extract target text features and corresponding target feature values;
[0128] Substituting the target text feature and the target feature value into a preset prompt word template to generate an initial prompt word;
[0129] The initial prompt words are optimized by using a context-aware algorithm, the semantic expression is adjusted and the semantic weight of the keywords is enhanced, and the target prompt words of the product output model are generated.
[0130] During specific implementation, the specific implementation process of the steps in this embodiment is roughly as follows:
[0131] Step 1: User portrait analysis and feature extraction
[0132] 1. Parse the user profile: Parse the user profile to extract the target text features and corresponding target feature values. The user profile is typically a structured data format (such as a JSON object) that contains information such as the user's age, gender, occupation, income, product requirement type, risk appetite, and budget.
[0133] 2. Extract key features: Extract text features related to product output and their corresponding feature values from user portraits. For example:
[0134] Product demand type (e.g., "health insurance")
[0135] Risk preference (e.g., "low risk")
[0136] Budget range (e.g., “5,000 yuan / year”)
[0137] Age (e.g., "30 years old")
[0138] Gender (such as "male")
[0139] Occupation (such as "software engineer")
[0140] Result: A set of target text features and their corresponding target feature values are obtained, which is ready for generating the initial prompt words.
[0141] Step 2: Generate initial prompt words
[0142] 1. Select a prompt word template: Select an appropriate prompt word template based on the preset prompt word generation strategy. The prompt word template is a natural language sentence structure used to describe the user's needs. For example:
[0143] Template 1: "Recommend an insurance product suitable for [age] years old, [sex], occupation [occupation], annual income [income], who needs a [product requirement type], has a risk preference of [risk preference], and a budget of [budget]."
[0144] Template 2: "For [age] years old, [sex], occupation [occupation], annual income [income], recommend an insurance product with [product requirement type], budget within [budget], and risk preference [risk preference]."
[0145] 2. Fill in feature values: Substitute the extracted target text features and their corresponding target feature values into the preset prompt word template to generate the initial prompt word.
[0146] Result: A preliminary initial cue word is generated.
[0147] Step 3: Optimize prompt words
[0148] 1. Context-aware algorithm: Optimize the initial prompt word using a context-aware algorithm. This algorithm dynamically adjusts the semantic expression of the prompt word and enhances the semantic weight of the keyword based on the contextual information of the user profile.
[0149] 2. Adjust semantic expression:
[0150] Simplify sentence structure: Make sure your prompts are concise and clear, and avoid lengthy and complex expressions.
[0151] Correct grammatical errors: Ensure that the prompt words are grammatically correct and conform to natural language habits.
[0152] Optimize vocabulary selection: Use more accurate and natural vocabulary to express user needs.
[0153] 3. Enhance the semantic weight of keywords:
[0154] Keyword Preposition: Put the most important keywords at the front of the sentence to ensure they are recognized by the model first.
[0155] Clear qualifiers: Use clear qualifiers (e.g., “fits,” “falls within budget”) to emphasize key information.
[0156] Avoid redundant information: remove unnecessary repeated information and ensure the semantic weight of keywords.
[0157] Result: An optimized target prompt word is generated, which reflects user needs more naturally and accurately.
[0158] Through the above process, this embodiment can realize the generation from user portrait to target prompt words, ensuring that the product output model can provide accurate and personalized business product recommendation services.
[0159] Furthermore, in one embodiment, the product output method, wherein the target prompt word is used as input, a product output result is generated by the product output model, and the product output result is sent to the user terminal for display according to the display strategy, specifically comprises the steps of:
[0160] Inputting the target prompt word into the product output model to generate a product output result, wherein the product output result includes text content, image content, and audio content;
[0161] Formatting the text content, the image content, and the audio content;
[0162] The formatted text content, the image content, and the audio content are sent to the user terminal for display according to the display strategy.
[0163] Furthermore, the product output method, after sending the formatted text content, the image content, and the audio content to the user terminal for display according to the display strategy, further specifically includes the steps of:
[0164] collecting user interaction data on the user terminal regarding the text content, the image content, and the audio content, and receiving feedback information sent by the user terminal regarding the text content, the image content, and the audio content;
[0165] A data analysis tool is used to comprehensively analyze the interaction data and the feedback information, and the product output model is optimized according to the analysis results.
[0166] During specific implementation, the specific implementation process of the steps in this embodiment is roughly as follows:
[0167] Step 1: Input target prompt words and generate recommendation results
[0168] 1. Input target prompt words: Input the target prompt words into the product output model.
[0169] 2. Generate recommendation results: The product output model generates product output results based on the target prompt word. Recommendation results include various content forms:
[0170] Text content: detailed description of the business product, its features, coverage, etc.
[0171] Image content: promotional pictures of business products, charts, user interface screenshots, etc.
[0172] Audio content: voice introduction of business products, audio clips of product explanations, etc.
[0173] 3. Example:
[0174] Target prompt: "Recommend health insurance suitable for a 30-year-old male software engineer with a budget of less than 5,000 yuan and a low risk appetite."
[0175] Recommended results:
[0176] The text reads: "Health Care Plan A: Premium 4,800 yuan / year, covers major illnesses and medical expenses, suitable for users with a low risk preference."
[0177] Image content: Product promotional images, coverage charts.
[0178] Audio content: audio clips explaining the product.
[0179] Result: Generates product output results containing text content, image content, and audio content.
[0180] Step 2: Content formatting
[0181] 1. Text Formatting: Formatting and arranging text to ensure it is clear and easy to read on the user end. For example, using HTML tags to format text, add headings, paragraphs, lists, etc.
[0182] 2. Image content formatting: Optimize image content to ensure it displays well on different devices. For example, adjust image size, format (such as JPEG, PNG), and resolution.
[0183] 3. Audio content formatting: Encode and compress audio content to ensure smooth playback on the user end. For example, converting audio files to MP3 format to optimize audio quality.
[0184] Result: Formatted text, image, and audio content, ready for presentation.
[0185] Step 3: Display recommendation results
[0186] The formatted text, image, and audio content are displayed on the user side according to the display strategy specified by the user side. The display strategy may include:
[0187] Arrangement method: Sort by premium level, coverage size, etc.
[0188] Display quantity: Display the first 3 or 5 recommended results.
[0189] Display format: Display in the form of lists, cards, carousels, etc.
[0190] Result: The user end displays the product output results in a clear and easy-to-use form.
[0191] Step 4: Collect user interaction data and feedback information
[0192] 1. Interaction data collection: Collect user interaction data on text content, image content, and audio content on the user terminal, including:
[0193] Click behavior (such as clicking on a recommended result)
[0194] Dwell time (e.g., how long you stay on a recommended result)
[0195] Scrolling behavior (such as scrolling to view detailed information about recommended results)
[0196] Audio playback duration (such as the time the audio content is played)
[0197] 2. Feedback information collection: Receive feedback information sent by the user end regarding text content, image content, and audio content, including:
[0198] Ratings (such as user satisfaction ratings for recommendation results)
[0199] Comments (such as users' specific opinions on the recommended results)
[0200] Preference marks (such as user likes or dislikes for recommended results)
[0201] Results: The user interaction data and feedback information about text content, image content and audio content were collected.
[0202] Step 5: Data analysis and model optimization
[0203] 1. Data Analysis: Use data analysis tools (such as machine learning algorithms and data mining techniques) to conduct a comprehensive analysis of the collected interaction data and feedback information. The analysis may include:
[0204] User preferences for different recommendation results;
[0205] Click-through rate and product conversion rate of recommended results;
[0206] User satisfaction with recommendation results;
[0207] 2. Model optimization: Optimize the product output model based on the analysis results, such as adjusting the model's feature weights, improving the recommendation algorithm, etc.
[0208] Results: Through data analysis and model optimization, the accuracy of the recommendation system and user experience are improved.
[0209] Through the above process, this embodiment can realize the complete flow from target prompt word input to product output result display, and can further optimize the product output model through user feedback and interaction data, thereby ensuring the continuous improvement and optimization of the product output system.
[0210] As can be seen from the above method embodiments, the product output method provided by the present invention includes: constructing a product output model through transfer learning and fine-tuning training based on a pre-trained large language model; receiving user information, preference data and display strategy input by the user terminal, and constructing a user portrait based on the user information and the preference data; constructing the target prompt word of the product output model based on the user portrait and the preset prompt word generation strategy; using the target prompt word as input, generating a product output result through the product output model, and sending the product output result to the user terminal for display according to the display strategy. In this way, the method of the present invention can realize accurate and personalized business product recommendation services.
[0211] It should be understood that although the present application provides method operation steps as described in the embodiments or flowcharts, more or fewer operation steps may be included based on conventional or non-creative work, and these operation steps are not necessarily performed in the order of the embodiments or flowcharts. The order of steps listed in the embodiments or flowcharts is only one way of executing the steps among many steps and does not represent the only execution order. It should be noted that there is not necessarily a certain order between the above steps. Those of ordinary skill in the art can understand from the description of the embodiments of the present invention that in different embodiments, the above steps may have different execution orders, that is, they may be executed in parallel, or they may be executed in an interchangeable manner, etc. Moreover, at least a portion of the steps in the embodiments or flowcharts may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but may be executed at different times. The execution order of these sub-steps or stages is not necessarily to be performed in sequence, but may be executed in turn, alternately or synchronously with other steps or at least a portion of the sub-steps or stages of other steps.
[0212] Based on the above method embodiment, please refer to Figure 3 Another embodiment of the present invention further provides a product output device, wherein the device includes:
[0213] Model building module 11, used to build a product output model based on the pre-trained large language model through transfer learning and fine-tuning training;
[0214] A portrait construction module 12 is configured to receive user information, preference data, and display strategies input by a user terminal, and to construct a user portrait based on the user information and preference data;
[0215] A prompt word construction module 13 is used to construct target prompt words of the product output model according to the user portrait and a preset prompt word generation strategy;
[0216] The result generation module 14 is configured to take the target prompt word as input, generate a product output result through the product output model, and send the product output result to the user terminal for display according to the display strategy.
[0217] Furthermore, in one embodiment, the product output device, wherein the model building module 11 is specifically used to:
[0218] Collect product recommendation data related to the target business products, and clean, label, and extract features from the product recommendation data to obtain a training data set;
[0219] According to a preset training data partitioning strategy, the training data set is divided into a migration data set, a model training set, a model validation set, and a model test set;
[0220] Performing transfer learning on the pre-trained large language model using the migration dataset, and performing fine-tuning training on the large language model after transfer learning using the model training set;
[0221] Evaluating the large language model after fine-tuning training using the model validation set, and optimizing the large language model according to the evaluation results;
[0222] The optimized large language model is tested on the model test set, and when the large language model meets the recommendation performance requirements of the target business product, a product output model is generated.
[0223] Furthermore, in one embodiment, the product output device, wherein the portrait construction module 12 is specifically configured to:
[0224] Receive user information, preference data, and display strategy input by a user, wherein the user information includes the user's age, gender, occupation, and income; the preference data includes the user's product demand type, risk preference, and budget range; and the display strategy includes the arrangement, display quantity, and display format of the product output results;
[0225] Preprocessing the user information and the preference data, including removing irrelevant characters and unifying text formats;
[0226] Using natural language processing technology to perform word segmentation and feature extraction on the preprocessed user information and preference data to obtain text features and corresponding feature values of the user information and preference data;
[0227] A user profile of the user is constructed based on the text features and the corresponding feature values.
[0228] Furthermore, the product output device, wherein the use of natural language processing technology to perform word segmentation and feature extraction on the pre-processed user information and preference data to obtain text features and corresponding feature values of the user information and preference data, specifically includes:
[0229] Using natural language processing technology to perform word segmentation on the pre-processed user information and preference data, dividing the text of the user information and preference data into multiple independent vocabulary units, and generating word segmentation results;
[0230] Through keyword extraction and entity recognition, feature extraction is performed on the word segmentation results to obtain text features and corresponding feature values therein;
[0231] Each of the text features is associated with the feature value corresponding to it, so that each of the text features is in one-to-one correspondence with the feature value corresponding to it.
[0232] Furthermore, in one embodiment, the product output device, wherein the prompt word construction module 13 is specifically used to:
[0233] Analyze the user profile to extract target text features and corresponding target feature values;
[0234] Substituting the target text feature and the target feature value into a preset prompt word template to generate an initial prompt word;
[0235] The initial prompt words are optimized by using a context-aware algorithm, the semantic expression is adjusted and the semantic weight of the keywords is enhanced, and the target prompt words of the product output model are generated.
[0236] Furthermore, in one embodiment, the product output device, wherein the result generation module 14 is specifically configured to:
[0237] Inputting the target prompt word into the product output model to generate a product output result, wherein the product output result includes text content, image content, and audio content;
[0238] Formatting the text content, the image content, and the audio content;
[0239] The formatted text content, the image content, and the audio content are sent to the user terminal for display according to the display strategy.
[0240] Furthermore, the product output device, after sending the formatted text content, the image content, and the audio content to the user terminal for display according to the display strategy, further specifically includes:
[0241] collecting user interaction data on the user terminal regarding the text content, the image content, and the audio content, and receiving feedback information sent by the user terminal regarding the text content, the image content, and the audio content;
[0242] A data analysis tool is used to comprehensively analyze the interaction data and the feedback information, and the product output model is optimized according to the analysis results.
[0243] It should be noted that, in the embodiment of the device of the present invention, the information interaction, execution process and other contents between the above modules are based on the same concept as the embodiment of the method of the present invention. Their specific functions and technical effects can be found in the aforementioned method embodiment part and will not be repeated here.
[0244] Based on the above method embodiment, another embodiment of the present invention further provides a computer device, which can be a server, and its internal structure diagram can be as follows: Figure 4 As shown. The computer device includes a processor, memory, network interface and database connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the functions or steps on the service side of the product output method in any of the above method embodiments are implemented.
[0245] Based on the above method embodiment, another embodiment of the present invention further provides a computer device, which can be a client, and its internal structure diagram can be as follows: Figure 5 As shown. The computer device includes a processor, memory, network interface, display screen, and input device connected via a system bus. 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 computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the functions or steps on the client side of the product output method in any of the above method embodiments are implemented.
[0246] Those skilled in the art will understand that Figure 4 and Figure 5 The structural diagram shown in the figure is only a schematic diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied. The specific computer device may include more components than shown in the figure, or combine certain components, or have a different component arrangement.
[0247] The processor referred to herein may be a CPU, other general-purpose processors, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor, or any conventional processor, etc.
[0248] The memory includes a readable storage medium, an internal memory, etc., wherein the internal memory can be the memory of a computer device, and the internal memory provides an environment for the operation of the operating system and computer-readable instructions in the readable storage medium. The readable storage medium can be a hard disk of the computer device, and in other embodiments, it can also be an external storage device of the computer device, for example, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the computer device. Furthermore, the memory can also include both an internal storage unit of the computer device and an external storage device. The memory is used to store an operating system, an application program, a boot loader (BootLoader), data, and other programs, such as the program code of a computer program. The memory can also be used to temporarily store data that has been output or is about to be output.
[0249] Based on the above method embodiments, another embodiment of the present invention further provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the product output method described in any of the above method embodiments. The computer-readable storage medium may be non-volatile or volatile.
[0250] It should be noted that the above-mentioned functions or steps that can be implemented by computer-readable storage media or computer devices, and the technical effects brought about by the functions / steps, can be found in the relevant descriptions in the aforementioned method embodiments. To avoid repetition, they will not be described one by one here.
[0251] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. 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. By way of illustration and not limitation, RAM is available in various 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). The disclosed memory components or memories of the operating environments described herein are intended to comprise one or more of these and / or any other suitable types of memory.
[0252] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, in the embodiment of the device of the present invention, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual application, 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. The functional units and modules in the embodiment can be integrated into a processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other and are not used to limit the scope of protection of the present invention. The specific working process of the units and modules in the above-mentioned device can refer to the corresponding process in the above-mentioned method embodiment, which will not be repeated here. If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium.
[0253] Those skilled 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. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present invention.
[0254] In the embodiments provided by the present invention, it should be understood that the disclosed apparatus / computer equipment and methods can be implemented in other ways. For example, the apparatus / computer equipment embodiments described above are merely illustrative. For example, the division of modules or units is merely a logical function division. In actual implementation, there may be other division methods, 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 through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.
[0255] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.
[0256] It should be noted that if software tools or components other than those of the Company appear in the embodiments of this application, they are merely for illustration and do not represent actual use. The above embodiments are intended only to illustrate the technical solutions of the present invention, not 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 with 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 various embodiments of the present invention, and should all be included in the scope of protection of the present invention.
Claims
1. A product output method, characterized in that: include: Based on the pre-trained large language model, the product output model is constructed through transfer learning and fine-tuning training; Receive user information, preference data, and display strategy input by the user terminal, and build a user profile based on the user information and preference data; Constructing target prompt words for the product output model based on the user portrait and a preset prompt word generation strategy; The target prompt word is used as input, a product output result is generated through the product output model, and the product output result is sent to the user terminal for display according to the display strategy.
2. The product output method according to claim 1, characterized in that: The product output model is constructed through transfer learning and fine-tuning training based on the pre-trained large language model, including: Collect product recommendation data related to the target business products, and clean, label, and extract features from the product recommendation data to obtain a training data set; According to a preset training data partitioning strategy, the training data set is divided into a migration data set, a model training set, a model validation set, and a model test set; Performing transfer learning on the pre-trained large language model using the migration dataset, and performing fine-tuning training on the large language model after transfer learning using the model training set; Evaluating the large language model after fine-tuning training using the model validation set, and optimizing the large language model according to the evaluation results; The optimized large language model is tested on the model test set, and when the large language model meets the recommendation performance requirements of the target business product, a product output model is generated.
3. The product output method according to claim 1, characterized in that: The receiving of user information, preference data, and display strategy input by the user terminal, and constructing a user profile based on the user information and preference data, includes: Receive user information, preference data, and display strategy input by a user, wherein the user information includes the user's age, gender, occupation, and income; the preference data includes the user's product demand type, risk preference, and budget range; and the display strategy includes the arrangement, display quantity, and display format of the product output results; Preprocessing the user information and the preference data, including removing irrelevant characters and unifying text formats; Using natural language processing technology to perform word segmentation and feature extraction on the preprocessed user information and preference data to obtain text features and corresponding feature values of the user information and preference data; A user profile of the user is constructed based on the text features and the corresponding feature values.
4. The product output method according to claim 3, characterized in that: The using of natural language processing technology to perform word segmentation and feature extraction on the pre-processed user information and preference data to obtain text features and corresponding feature values of the user information and preference data includes: Using natural language processing technology to perform word segmentation on the pre-processed user information and preference data, dividing the text of the user information and preference data into multiple independent vocabulary units, and generating word segmentation results; Through keyword extraction and entity recognition, feature extraction is performed on the word segmentation results to obtain text features and corresponding feature values therein; Each of the text features is associated with the feature value corresponding to it, so that each of the text features is in one-to-one correspondence with the feature value corresponding to it.
5. The product output method according to claim 1, characterized in that: The step of constructing the target prompt words of the product output model according to the user portrait and the preset prompt word generation strategy includes: Analyze the user profile to extract target text features and corresponding target feature values; Substituting the target text feature and the target feature value into a preset prompt word template to generate an initial prompt word; The initial prompt words are optimized by using a context-aware algorithm, the semantic expression is adjusted and the semantic weight of the keywords is enhanced, and the target prompt words of the product output model are generated.
6. The product output method according to claim 1, characterized in that: The step of taking the target prompt word as input, generating a product output result through the product output model, and sending the product output result to the user terminal for display according to the display strategy includes: Inputting the target prompt word into the product output model to generate a product output result, wherein the product output result includes text content, image content, and audio content; Formatting the text content, the image content, and the audio content; The formatted text content, the image content, and the audio content are sent to the user terminal for display according to the display strategy.
7. The product output method according to claim 6, characterized in that: After sending the formatted text content, the image content, and the audio content to the user terminal for display according to the display strategy, the method further includes: collecting user interaction data on the user terminal regarding the text content, the image content, and the audio content, and receiving feedback information sent by the user terminal regarding the text content, the image content, and the audio content; A data analysis tool is used to comprehensively analyze the interaction data and the feedback information, and the product output model is optimized according to the analysis results.
8. A product output device, characterized in that: include: The model building module is used to build a product output model through transfer learning and fine-tuning training based on a pre-trained large language model; A portrait building module is used to receive user information, preference data and display strategy input by the user terminal, and build a user portrait based on the user information and preference data; A prompt word construction module, configured to construct target prompt words for the product output model based on the user profile and a preset prompt word generation strategy; The result generation module is configured to take the target prompt word as input, generate a product output result through the product output model, and send the product output result to the user terminal for display according to the display strategy.
9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the product output method according to any one of claims 1 to 7 is implemented.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the product output method according to any one of claims 1 to 7 is implemented.