Information push strategy determination method and device, computer equipment and readable storage medium
By acquiring user and business personnel characteristics, and using business personnel matching models and large language models to generate product recommendation scripts, the problem of inaccurate information push to isolated users was solved, personalized information push was achieved, and user churn was reduced.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-02
- Publication Date
- 2026-03-06
AI Technical Summary
When business personnel have high turnover, it is difficult to deliver accurate information to isolated users, and new business personnel may not be able to meet user needs, resulting in poor information delivery effectiveness.
By acquiring the characteristics of target users and business personnel, a pre-trained business personnel matching model is used to identify target business personnel. Target prompts are generated by combining user characteristics and product characteristics. The input is then used to generate product recommendation scripts using a large language model, and a target information push strategy is formulated.
It enables precise information delivery to lonely users, improving the accuracy of information delivery and reducing user churn.
Smart Images
Figure CN121616366A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of computer technology, and in particular to a method, apparatus, computer device, computer-readable storage medium, and computer program product for determining an information push strategy. Background Technology
[0002] In the course of business operations, the high turnover rate of sales personnel who interact with users is a long-standing pain point in the industry. As sales personnel leave, the users who were originally provided with follow-up services gradually become isolated users who lack targeted maintenance because they lose their dedicated service personnel.
[0003] In related technologies, companies typically assign new business personnel to lonely users; however, on the one hand, the new business personnel lack understanding of lonely users, and on the other hand, there may be a mismatch between the user's needs and the new business personnel's areas of expertise, making it difficult to accurately push information to lonely users. Summary of the Invention
[0004] Therefore, it is necessary to address the aforementioned technical problem of difficulty in accurately pushing information to lonely users by providing a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for determining an information push strategy that can accurately push information to lonely users.
[0005] Firstly, this application provides a method for determining an information push strategy, including:
[0006] Obtain user characteristics of the target user and business personnel characteristics of multiple business personnel;
[0007] The user features and the business personnel features are input into a pre-trained business personnel matching model to identify the target business personnel who match the target user from among the multiple business personnel.
[0008] By combining the initial prompt information, the user characteristics, and the product characteristics of the product to be recommended, the target prompt information corresponding to the target user is obtained; both the initial prompt information and the target prompt information are used to prompt the generation of product recommendation scripts.
[0009] The target prompt information is input into the large language model to obtain the target product recommendation script for the target user.
[0010] Based on the target business personnel and the target product recommendation script, a target information push strategy is obtained for the target user.
[0011] In one embodiment, the step of inputting the user features and the business personnel features into a pre-trained business personnel matching model to determine the target business personnel matching the target user from among a plurality of business personnel includes:
[0012] The user characteristics and the business personnel characteristics of each business personnel are concatenated into a single input data, resulting in multiple input data;
[0013] The input data is fed into the business personnel matching model to obtain the recommendation success probability information of each business personnel for the target user.
[0014] The target business personnel are obtained from among the multiple business personnel whose corresponding recommendation success probability information meets the preset recommendation success probability conditions.
[0015] In one embodiment, the number of user features is multiple;
[0016] The combination of the initial prompt information, the user characteristics, and the product characteristics of the product to be recommended yields the target prompt information corresponding to the target user, including:
[0017] Target user features are determined from the user features of the target user; the target user features are user features whose importance meets a preset importance condition.
[0018] The target prompt information is generated by combining the initial prompt information, the target user characteristics, and the product characteristics of the product to be recommended.
[0019] In one embodiment, the output information of the business personnel matching model includes at least the recommendation success probability information of each business personnel corresponding to the target user and the target user characteristics of the target user.
[0020] In one embodiment, the business personnel matching model is trained in the following manner:
[0021] The system acquires user characteristics of multiple sample users, business personnel characteristics of multiple sample business personnel, and historical recommendation information between each sample user and each sample business personnel. The historical recommendation information is used to characterize whether there is a successful recommendation behavior between the corresponding sample business personnel and the corresponding sample user.
[0022] Each sample business personnel's corresponding business personnel feature and a historical recommendation information are concatenated into a first training data, resulting in multiple first training data;
[0023] Each sample user's corresponding user feature and a historical recommendation information are concatenated into a second training data, resulting in multiple second training data;
[0024] The business personnel matching model to be trained is iteratively trained based on multiple sets of the first training data and multiple sets of the second training data until the training termination condition is met, thereby obtaining the business personnel matching model.
[0025] In one embodiment, after obtaining the target information push strategy for the target user, the method further includes:
[0026] After obtaining the target information push strategy for the target user, the method further includes:
[0027] The target product recommendation script in the target information push strategy is sent to the terminal of the target business personnel in the target information push strategy; the target business personnel are used to recommend products to the target users based on the target product recommendation script.
[0028] Secondly, this application also provides an information push strategy determination device, comprising:
[0029] The feature acquisition module is used to acquire user features of the target user and business personnel features of multiple business personnel;
[0030] The personnel matching module is used to input the user features and the business personnel features into a pre-trained business personnel matching model, and to determine the target business personnel who match the target user from among the multiple business personnel.
[0031] The information generation module is used to combine the initial prompt information, the user characteristics, and the product characteristics of the product to be recommended to obtain the target prompt information corresponding to the target user; both the initial prompt information and the target prompt information are used to prompt the generation of product recommendation scripts;
[0032] The script generation module is used to input the target prompt information into the large language model to obtain the target product recommendation script for the target user;
[0033] The strategy determination module is used to obtain a target information push strategy for the target user based on the target business personnel and the target product recommendation script.
[0034] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:
[0035] Obtain user characteristics of the target user and business personnel characteristics of multiple business personnel;
[0036] The user features and the business personnel features are input into a pre-trained business personnel matching model to identify the target business personnel who match the target user from among the multiple business personnel.
[0037] By combining the initial prompt information, the user characteristics, and the product characteristics of the product to be recommended, the target prompt information corresponding to the target user is obtained; both the initial prompt information and the target prompt information are used to prompt the generation of product recommendation scripts.
[0038] The target prompt information is input into the large language model to obtain the target product recommendation script for the target user.
[0039] Based on the target business personnel and the target product recommendation script, a target information push strategy is obtained for the target user.
[0040] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:
[0041] Obtain user characteristics of the target user and business personnel characteristics of multiple business personnel;
[0042] The user features and the business personnel features are input into a pre-trained business personnel matching model to identify the target business personnel who match the target user from among the multiple business personnel.
[0043] By combining the initial prompt information, the user characteristics, and the product characteristics of the product to be recommended, the target prompt information corresponding to the target user is obtained; both the initial prompt information and the target prompt information are used to prompt the generation of product recommendation scripts.
[0044] The target prompt information is input into the large language model to obtain the target product recommendation script for the target user.
[0045] Based on the target business personnel and the target product recommendation script, a target information push strategy is obtained for the target user.
[0046] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:
[0047] Obtain user characteristics of the target user and business personnel characteristics of multiple business personnel;
[0048] The user features and the business personnel features are input into a pre-trained business personnel matching model to identify the target business personnel who match the target user from among the multiple business personnel.
[0049] By combining the initial prompt information, the user characteristics, and the product characteristics of the product to be recommended, the target prompt information corresponding to the target user is obtained; both the initial prompt information and the target prompt information are used to prompt the generation of product recommendation scripts.
[0050] The target prompt information is input into the large language model to obtain the target product recommendation script for the target user.
[0051] Based on the target business personnel and the target product recommendation script, a target information push strategy is obtained for the target user.
[0052] The aforementioned information push strategy determination method, apparatus, computer equipment, computer-readable storage medium, and computer program product, on the one hand, can identify the target business personnel matching the target user from multiple business personnel by using the user characteristics of the target user and the business personnel characteristics of the business personnel; on the other hand, by using the user characteristics of the target user, it can generate target prompt information for generating product recommendation scripts for the target user, and thus generate target product recommendation scripts for the target user. Based on the target business personnel and target product recommendation scripts, personalized information push can be carried out for the user, thus achieving accurate information push to isolated users and improving the accuracy of information push. Attached Figure Description
[0053] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0054] Figure 1 This is an application environment diagram of the information push strategy determination method in one embodiment;
[0055] Figure 2 This is a flowchart illustrating the information push strategy determination method in one embodiment;
[0056] Figure 3 This is a flowchart illustrating the steps of training a business personnel matching model in one embodiment;
[0057] Figure 4 This is a flowchart illustrating an insurance marketing assistance method that integrates salesperson matching and script generation in one embodiment.
[0058] Figure 5 This is a structural block diagram of an information push strategy determination device in one embodiment;
[0059] Figure 6 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation
[0060] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0061] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.
[0062] It should be noted that the user information (including but not limited to user information and business personnel information) and data (including but not limited to user data and business personnel data used for analysis, storage, and display by users and business personnel) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of related data must comply with relevant regulations. In practical applications, business personnel can refuse or easily refuse product recommendation scripts pushed to them, and users can refuse or easily refuse pushed products or information.
[0063] The information push strategy determination method provided in this application embodiment can be applied to, for example, Figure 1In the application environment shown, server 102 communicates with terminal 104 via a network; the data storage system can store the data that server 102 needs to process, and the data storage system can be integrated on server 102 or placed on the cloud or other network servers. Furthermore, server 102 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services; terminal 104 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, drones, low-altitude aircraft, IoT devices, and portable wearable devices. IoT devices can be smart speakers, smart TVs, smart air conditioners, smart vehicle devices, projection devices, etc. Portable wearable devices can be smartwatches, smart bracelets, head-mounted devices, etc. Head-mounted devices can be virtual reality (VR) devices, augmented reality (AR) devices, smart glasses, etc.
[0064] For example, server 102 first obtains the user characteristics of the target user and the business personnel characteristics of multiple business personnel; then, it inputs the user characteristics and business personnel characteristics into a pre-trained business personnel matching model to identify the target business personnel matching the target user among the multiple business personnel; next, it combines the initial prompt information, user characteristics, and product characteristics of the product to be recommended to obtain the target prompt information corresponding to the target user; both the initial prompt information and the target prompt information are used to prompt the generation of product recommendation scripts; then, it inputs the target prompt information into a large language model to obtain the target product recommendation scripts for the target user; finally, based on the target business personnel and the target product recommendation scripts, it obtains the target information push strategy for the target user and sends the target product recommendation scripts to the target business personnel's terminal 104.
[0065] In some embodiments, such as Figure 2 As shown, a method for determining an information push strategy is provided, which is applied to... Figure 1 Taking the server in the example, the following steps are included:
[0066] Step S202: Obtain the user characteristics of the target user and the business personnel characteristics of multiple business personnel.
[0067] The target users are lonely users; lonely users are those who have lost their dedicated service personnel due to the departure of the original service personnel.
[0068] The number of target users is at least one.
[0069] In practical applications, users refer to the company's customers, and business personnel refer to the company's marketing staff.
[0070] In this step, the server first identifies at least one target user and multiple business personnel, and then obtains the user information of each target user and the business personnel information of the multiple business personnel. Next, for each target user, feature extraction processing is performed on the user information of the target user to obtain the user features of the target user, and for each business personnel, feature extraction processing is performed on the business personnel information of the business personnel to obtain the business personnel features of the business personnel.
[0071] For example, user information includes at least basic user information and historical product purchase behavior information; basic user information includes at least age, gender, and region; historical product purchase behavior information includes at least the type and quantity of products purchased, total cost of products purchased, and average cost of products purchased within the first statistical time interval. Taking the insurance industry as an example, historical product purchase behavior information includes the type of insurance policies purchased, the number of insurance policies purchased, the total premium of the policies, and the average premium of the policies purchased within the first statistical time interval.
[0072] For example, sales personnel information includes at least basic information, professional information, performance information, and visit information. Basic information includes at least gender, age, and education level. Professional information includes at least the length of time the sales personnel have been with the company, the length of time they have held contracts, and their current job level. Performance information includes at least the number of products sold, the amount of money sold, commission income, and the number of products sold per customer within the second statistical time period. For example, in the insurance industry, performance information includes the number of policies, premium amount, commission income, and the number of policies sold per customer within the second statistical time period. Visit information includes at least the number of users visited, the number of visits, and the number of visits per customer within the third statistical time period.
[0073] Step S204: Input the user features and business personnel features into the pre-trained business personnel matching model to identify the target business personnel who match the target user from among multiple business personnel.
[0074] The business personnel matching model is trained based on a neural network model or a machine learning model; in practical applications, the business personnel matching model is trained based on the XGBoost (Extreme Gradient Boosting Tree) model.
[0075] In this step, for each target user, the server inputs the user characteristics of the target user and the business personnel characteristics of each business person into a pre-trained business personnel matching model. Through the matching process of the business personnel matching model, the target business person matching the target user is determined from multiple business personnel.
[0076] For example, the number of target business personnel is at least one.
[0077] Step S206: Combine the initial prompt information, user characteristics, and product characteristics of the product to be recommended to obtain the target prompt information corresponding to the target user.
[0078] The initial prompts and target prompts are both used to generate product recommendation messages, and the prompts are prompt words from the large language model.
[0079] The number of products to be recommended is at least one.
[0080] In this embodiment, the server obtains the product information of each product to be recommended by the enterprise, and performs feature extraction processing on the product information of each product to be recommended to obtain the product features of each product to be recommended; then, for each target user, the server injects the user's user features and the product features of each product to be recommended into the initial prompt information to generate the target prompt information corresponding to the target user.
[0081] For example, product information includes at least the product name, product content, product type, and payment method. Taking the insurance industry as an example, the product is an insurance product, and the product information includes the insurance name, the core insurance coverage, the insurance type, and the payment method.
[0082] Step S208: Input the target prompt information into the large language model to obtain the target product recommendation script for the target user.
[0083] In this step, for each target user, the server inputs the target prompt information corresponding to that target user into the large language model, so as to generate a target product recommendation script for the target user through the large language model.
[0084] Step S210: Based on the target business personnel and the target product recommendation script, obtain the target information push strategy for the target users.
[0085] In this step, for each target user, the server obtains a target information push strategy based on the target business personnel and target product recommendation script corresponding to that target user; furthermore, the server can also send the target product recommendation script corresponding to that target user to the target business personnel corresponding to that target user.
[0086] In the aforementioned method for determining the information push strategy, on the one hand, the server can identify the target business personnel matching the target user from multiple business personnel by using the user characteristics of the target user and the business personnel characteristics of the business personnel. On the other hand, the server can generate target prompt information for generating product recommendation scripts for the target user by using the user characteristics of the target user, and then generate target product recommendation scripts for the target user. Based on the target business personnel and target product recommendation scripts, personalized information push can be carried out for the user, thus realizing accurate information push to lonely users, improving the accuracy of information push, and thus reducing the churn of lonely users.
[0087] In some embodiments, step S204 above, which involves inputting user features and business personnel features into a pre-trained business personnel matching model to determine the target business personnel matching the target user from among multiple business personnel, includes the following steps: concatenating user features and business personnel features of each business personnel into a single input data set to obtain multiple input data sets; inputting each input data set into the business personnel matching model to obtain the recommendation success probability information for each business personnel corresponding to the target user; and obtaining the target business personnel based on the business personnel whose corresponding recommendation success probability information satisfies the preset recommendation success probability condition among the multiple business personnel.
[0088] Among them, the recommendation success probability information is the recommendation success probability value, which represents the likelihood that business personnel will successfully recommend products to users.
[0089] The preset success probability condition can be either a preset probability threshold or a preset ranking.
[0090] In this embodiment, for each target user, the server first concatenates the user characteristics of the target user and the business personnel characteristics of each business person into a single input data, thereby obtaining multiple input data for the business personnel matching model. For example, the server performs a Cartesian product concatenation on the user characteristics and the business personnel characteristics to obtain a single input data. Then, the server inputs each input data into the business personnel matching model. Through the learning of each input data by the business personnel matching model, the server predicts the recommendation success probability information for each business person corresponding to the target user. Next, the server determines the business persons whose recommendation success probability value is greater than or equal to a preset probability threshold as the target business persons for the target user. Alternatively, the server sorts each target business person in descending order of recommendation success probability value and determines the business persons whose ranking is higher than or equal to a preset ranking as the target business persons for the target user. That is, the server selects the top k business persons as target business persons based on the recommendation success probability value.
[0091] In this embodiment, the server can identify the target business personnel that match the target user from among multiple business personnel by using the user characteristics of the target user and the business personnel characteristics of the business personnel.
[0092] In some embodiments, the number of user features is multiple.
[0093] Step S206 above, which combines the initial prompt information, user characteristics, and product characteristics of the product to be recommended to obtain the target prompt information corresponding to the target user, includes the following steps: determining the target user characteristics from the various user characteristics of the target user; combining the initial prompt information, target user characteristics, and product characteristics of the product to be recommended to generate the target prompt information.
[0094] The initial prompts define the roles and tasks of the large language model.
[0095] Among them, the target user characteristics are the user characteristics that meet the preset importance conditions; the target user characteristics of different target users can be different.
[0096] In this embodiment, for each target user, the server selects at least one important target user feature from multiple user features of the target user, and then injects each target user feature of the target user and the product features of each product to be recommended into the initial prompt information to generate the target prompt information corresponding to the target user.
[0097] The characteristics of target users include at least their needs, communication preferences, risk preferences, and product preferences.
[0098] For example, the server can inject the target user characteristics of the target user and the product characteristics of the products to be recommended into the initial prompt information through a large language model, and generate the target prompt information corresponding to the target user.
[0099] For example, the server can also inject output format requirements into the initial prompt message to generate the target prompt message.
[0100] In this embodiment, the server can generate corresponding target prompt information for the target user based on the target user's important target user characteristics.
[0101] In some embodiments, for each target user, the output information of the business personnel matching model includes at least the recommendation success probability information of each business personnel for that target user, and the target user characteristics of that target user.
[0102] In this embodiment, the server determines the target user characteristics of the target user based on the output information of the business personnel matching model.
[0103] In this embodiment, the server can determine the target user characteristics of the target user based on the output information of the business personnel matching model.
[0104] In some embodiments, such as Figure 3 As shown, the information push strategy determination method provided in this application also includes the following steps for training a business personnel matching model:
[0105] Step S302: Obtain user characteristics of multiple sample users, business personnel characteristics of multiple sample business personnel, and historical recommendation information between each sample user and each sample business personnel.
[0106] Step S304: Concatenate the business personnel features and historical recommendation information corresponding to each sample business personnel into a first training data point, thereby obtaining multiple first training data points.
[0107] Step S306: Concatenate the user features corresponding to each sample user and a historical recommendation information into a second training data set, thereby obtaining multiple second training data sets.
[0108] Step S308: Iteratively train the business personnel matching model to be trained based on multiple first training data and multiple second training data until the training termination condition is met, and obtain the business personnel matching model.
[0109] Historical recommendation information is used to characterize whether there have been successful recommendation interactions between the corresponding sample business personnel and the corresponding sample users. For example, whether the corresponding sample business personnel have successfully recommended a product to the corresponding sample user. In practical applications, each piece of historical recommendation information includes the user identifier of the corresponding sample user, the business personnel identifier of the corresponding sample business personnel, and a recommendation status message. The recommendation status message is used to characterize whether a successful recommendation interaction has occurred between the corresponding sample business personnel and the corresponding sample user.
[0110] In step S302, the server first identifies multiple sample users and multiple sample business personnel. Then, it obtains the user information of each sample user and the business personnel information of the multiple sample business personnel. Next, for each sample user, feature extraction processing is performed on the user information to obtain the user features of the sample user. Similarly, for each sample business personnel, feature extraction processing is performed on the business personnel information to obtain the business personnel features of the sample business personnel. Simultaneously, it determines whether there has been a successful recommendation behavior between each sample business personnel and each sample user. If the sample business personnel has successfully recommended a product to the sample user, the recommendation status information in the historical recommendation information between the sample user and the sample business personnel is determined to be a successful recommendation status. If the sample business personnel has not successfully promoted a product to the sample user, the recommendation status information in the historical recommendation information between the sample user and the sample business personnel is determined to be a failed recommendation status.
[0111] In steps S304 and S306, the user features of each sample user include at least the user identifier of the sample user, the business personnel features of each sample business person include at least the business personnel identifier of the sample business person, and each historical recommendation information includes the user identifier of the corresponding sample user, the business personnel identifier of the corresponding sample business person, and a recommendation status message; for each historical recommendation information, the server concatenates the historical recommendation information and the business personnel features of the sample business person corresponding to the historical recommendation information into a first training data, thereby obtaining multiple first training data; and concatenates the historical recommendation information and the user features of the sample user corresponding to the historical recommendation information into a second training data, thereby obtaining multiple second training data.
[0112] In step S308, the server uses multiple sets of first training data and multiple sets of second training data as training datasets to iteratively train the business personnel matching model to be trained until the training termination condition is met. The model that meets the training termination condition is then used as the business personnel matching model.
[0113] For example, the server uses an XGBoost model as the business personnel matching model to be trained. The server divides the training dataset into a training set and a test set, trains multiple XGBoost models based on the dataset, and tests the multiple trained XGBoost models based on the test set. Common model evaluation criteria such as AUC (Area Under the ROC Curve) and KS (Kolmogorov-Smirnov Statistic) are selected from the multiple trained XGBoost models to choose the optimal XGBoost model as the business personnel matching model.
[0114] In this embodiment, the server can train a business personnel matching model based on the historical recommendation behavior between sample users and sample business personnel, as well as between sample users and sample business personnel.
[0115] In some embodiments, after obtaining the target information push strategy for the target user in the above steps, the following is also included: sending the target product recommendation script in the target information push strategy to the terminal of the target business personnel in the target information push strategy.
[0116] Among them, the target sales personnel are used to recommend products to target users based on the target product recommendation script.
[0117] In this embodiment, for each target information push strategy, the server sends the target product recommendation script in the target information push strategy to the terminal of the target business personnel in the target information push strategy, so that the target business personnel can recommend products to the target users corresponding to the target information push strategy based on the target product recommendation script.
[0118] In this embodiment, the target business personnel matched with the target user can recommend products to the target user based on the target product recommendation script corresponding to the target user, thereby realizing personalized product recommendations for users and reducing the churn of lonely users.
[0119] To more clearly illustrate the information push strategy determination method provided in the embodiments of this application, the following specific embodiment is used to describe the information push strategy determination method in detail. However, it should be understood that the embodiments of this application are not limited thereto. Figure 4 As shown, in some embodiments, this application also provides an insurance marketing assistance method that integrates salesperson matching and script generation, specifically including the following steps:
[0120] 1. Match the best marketer to lonely users.
[0121] The training phase of the model involves: acquiring customer characteristics of multiple sample lonely customers, salesperson characteristics of multiple sample marketers, and historical marketing information between each sample lonely customer and each sample marketer; historical marketing information is used to characterize whether there were successful marketing behaviors between the corresponding sample marketer and the corresponding sample lonely customer; concatenating the salesperson characteristics and historical marketing information corresponding to each sample marketer into a first training data set, resulting in multiple first training data sets; concatenating the customer characteristics and historical marketing information corresponding to each sample lonely customer into a second training data set, resulting in multiple second training data sets; iteratively training the salesperson matching model to be trained based on the multiple first training data sets and the multiple second training data sets until the training termination condition is met, thus obtaining the salesperson matching model.
[0122] The inference phase of the model involves: acquiring the customer characteristics of the target lonely customer and the marketing characteristics of multiple marketing agents; concatenating the customer characteristics and the marketing characteristics of each marketing agent into a single input data set to obtain multiple input data sets; inputting each input data set into the marketing agent matching model to obtain the marketing success probability information for each marketing agent corresponding to the target lonely customer; and identifying the marketing agent whose corresponding marketing success probability information meets the preset marketing success probability conditions among the multiple marketing agents to obtain the target marketing agent.
[0123] 2. Generate product marketing scripts for lonely customers.
[0124] In addition to the marketing success probability information for each marketer corresponding to the target lonely customer, the output information of the marketer matching model also includes the more important customer characteristics of the target lonely customer. These customer characteristics constitute the customer profile of the target lonely customer. The initial prompt words, the customer profile of the target lonely customer and the product characteristics of the product to be recommended are combined to generate target prompt words. The target prompt words are input into the large language model so that the large language model can generate targeted marketing scripts for the target customer.
[0125] For example, the initial prompt word is as follows:
[0126] [Role] You are a senior insurance marketing expert who needs to generate product marketing scripts based on customer characteristics.
[0127] [Task] Generate a communication strategy based on the following dimensions: a 3-stage conversational structure: icebreaking -> needs discovery -> product solution.
[0128] The generation process of target prompt words can be modeled as the following formula:
[0129] Prompt += [Customer Profile] + [Product Features] + [Output Format]
[0130] The output format requirements are as follows: The final output should be in table format, and the table module should include icebreakers, product recommendations, and compliance tips.
[0131] In this embodiment, firstly, the first part (matching the best marketer to the lonely user) displays the marketer's sales ability and preferences based on the marketer's characteristics, and displays the lonely user's needs, communication preferences, and product preferences based on the user's characteristics. This identifies the most suitable marketer for the lonely user. Secondly, the second part (generating product marketing scripts for the lonely user) uses a large language model to generate precise product marketing scripts for the lonely user, since the most suitable marketer may not know the lonely user. Furthermore, the first part not only produces the best marketer but also characterizes the user's most important profile features. These profile features assist in the generation of prompts in the second part, achieving effective synergy between the two parts. Finally, the second part employs a dynamic prompt injection method. Because each user's user profile is different, personalized prompts can be dynamically generated, resulting in different product marketing scripts in a more reasonable and friendly way for the most suitable marketer to use.
[0132] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.
[0133] Based on the same inventive concept, this application also provides an information push strategy determination device for implementing the information push strategy determination method described above. The solution provided by this device is similar to the implementation described in the above method; therefore, the specific limitations in one or more information push strategy determination device embodiments provided below can be found in the limitations of the information push strategy determination method described above, and will not be repeated here.
[0134] In some embodiments, such as Figure 5 As shown, an information push strategy determination device is provided, including: a feature acquisition module 502, a personnel matching module 504, an information generation module 506, a script generation module 508, and a strategy determination module 510, wherein:
[0135] The feature acquisition module 502 is used to acquire the user features of the target user and the business personnel features of multiple business personnel.
[0136] The personnel matching module 504 is used to input user features and business personnel features into a pre-trained business personnel matching model to identify the target business personnel who match the target user from among multiple business personnel.
[0137] The information generation module 506 is used to combine initial prompt information, user characteristics, and product characteristics of the product to be recommended to obtain target prompt information corresponding to the target user; both the initial prompt information and the target prompt information are used to prompt the generation of product recommendation scripts.
[0138] The script generation module 508 is used to input target prompt information into the large language model to obtain target product recommendation scripts for target users.
[0139] The strategy determination module 510 is used to obtain a target information push strategy for target users based on the recommended scripts for target business personnel and target products.
[0140] In one embodiment, the personnel matching module 504 is further configured to concatenate user features and the business personnel features of each business personnel into a single input data set to obtain multiple input data sets; input each input data set into the business personnel matching model to obtain the recommendation success probability information of the target user corresponding to each business personnel; and obtain the target business personnel based on the business personnel whose corresponding recommendation success probability information meets the preset recommendation success probability conditions among the multiple business personnel.
[0141] In one embodiment, the number of user features is multiple.
[0142] The information generation module 506 is also used to determine the target user characteristics from the various user characteristics of the target user; the target customer characteristics are customer characteristics whose importance meets the preset importance conditions; and to generate target prompt information by combining the initial prompt information, the target user characteristics and the product characteristics of the product to be recommended.
[0143] In one embodiment, the output information of the business personnel matching model includes at least the recommendation success probability information of each business personnel corresponding to the target user and the target user characteristics of the target user.
[0144] In one embodiment, the information push strategy determination device further includes a model training module, used to acquire user features of multiple sample users, business personnel features of multiple sample business personnel, and historical recommendation information between each sample user and each sample business personnel; the historical recommendation information is used to characterize whether there is a successful recommendation behavior between the corresponding sample business personnel and the corresponding sample user; the business personnel features corresponding to each sample business personnel and a piece of historical recommendation information are concatenated into a first training data, to obtain multiple pieces of first training data; the user features corresponding to each sample user and a piece of historical recommendation information are concatenated into a second training data, to obtain multiple pieces of second training data; the business personnel matching model to be trained is iteratively trained according to the multiple pieces of first training data and the multiple pieces of second training data until the training termination condition is reached, to obtain the business personnel matching model.
[0145] In one embodiment, the information push strategy determination device further includes a script sending module, which sends the target product recommendation script in the target information push strategy to the terminal of the target business personnel in the target information push strategy; the target business personnel use the script to recommend products to the target users.
[0146] The modules in the aforementioned information push strategy determination device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.
[0147] In some embodiments, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 6 As shown, this computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media. The database stores user information, business personnel information, product information, and other data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements an information push strategy determination method.
[0148] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0149] In some embodiments, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.
[0150] In some embodiments, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.
[0151] In some embodiments, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.
[0152] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.
[0153] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.
[0154] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. An information push policy determination method, characterized by, The method comprises: obtaining user features of a target user and service personnel features of a plurality of service personnel; inputting the user features and the service personnel features into a pre-trained service personnel matching model to determine a target service personnel matched with the target user from the plurality of service personnel; combining initial prompt information, the user features and product features of a product to be recommended to obtain target prompt information corresponding to the target user; the initial prompt information and the target prompt information are both used for prompting product recommendation rhetoric; inputting the target prompt information into a large language model to obtain target product recommendation rhetoric for the target user; obtaining a target information push strategy for the target user based on the target service personnel and the target product recommendation rhetoric.
2. The method of claim 1, wherein, The inputting the user features and the service personnel features into a pre-trained service personnel matching model to determine a target service personnel matched with the target user from the plurality of service personnel comprises: concatenating the user features and the service personnel features of each of the service personnel into an input data to obtain a plurality of input data; inputting each of the input data into the service personnel matching model to obtain recommendation success probability information corresponding to the target user for each of the service personnel; obtaining the target service personnel according to service personnel whose corresponding recommendation success probability information meets a preset recommendation success probability condition from the plurality of service personnel.
3. The method of claim 1, wherein, The number of the user features is a plurality; The combining initial prompt information, the user features and product features of a product to be recommended to obtain target prompt information corresponding to the target user comprises: determining target user features from each of the user features of the target user; the target user features are user features whose importance meets a preset importance condition; combining the initial prompt information, the target user features and the product features of the product to be recommended to generate the target prompt information.
4. The method of claim 3, wherein, The output information of the service personnel matching model at least includes recommendation success probability information corresponding to the target user for each of the service personnel and the target user features of the target user.
5. The method according to any one of claims 1 to 4, characterized in that, The service personnel matching model is trained by the following way: obtaining user features of a plurality of sample users, service personnel features of a plurality of sample service personnel and historical recommendation information between each of the sample users and each of the sample service personnel; the historical recommendation information is used for representing whether there is a successful recommendation behavior between the corresponding sample service personnel and the corresponding sample user; concatenating the service personnel features of each of the sample service personnel and one of the historical recommendation information into a first training data to obtain a plurality of first training data; concatenating the user features of each of the sample users and one of the historical recommendation information into a second training data to obtain a plurality of second training data; The service personnel matching model to be trained is iteratively trained according to the plurality of pieces of first training data and the plurality of pieces of second training data until a training end condition is reached, to obtain the service personnel matching model.
6. The method according to any one of claims 1 to 4, characterized in that, After obtaining the target information push strategy for the target user, the method further includes: The target product recommendation script in the target information push strategy is sent to a terminal of the target service personnel in the target information push strategy; and the target service personnel is configured to recommend a product to the target user based on the target product recommendation script.
7. An information push strategy determination device, characterized in that, The apparatus includes: a feature acquisition module configured to acquire user features of a target user and service personnel features of a plurality of service personnel; a personnel matching module configured to input the user features and the service personnel features into a pre-trained service personnel matching model to determine a target service personnel matched with the target user from the plurality of service personnel; an information generation module configured to combine initial prompt information, the user features, and product features of a product to be recommended to obtain target prompt information corresponding to the target user; the initial prompt information and the target prompt information are both used to prompt generation of a product recommendation script; a script generation module configured to input the target prompt information into a large language model to obtain a target product recommendation script for the target user; a strategy determination module configured to obtain a target information push strategy for the target user based on the target service personnel and the target product recommendation script.
8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that, The processor executes the computer program to implement the steps of the method in any one of claims 1 to 6.
9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method in any one of claims 1 to 6.
10. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method in any one of claims 1 to 6.