User interest modeling method and system based on user explicit interest tag and high-intention behavior weighting

Abstract

The invention discloses a user interest modeling method and system based on user explicit interest labels and high-intention behavior weighting, and the method is characterized in that a user can independently select an interest label, and can add or cancel the interest label at any time; the system carries out matching recall on the platform information based on the interest label set to generate a candidate information set, and sorts the candidate information set to generate a recommendation list; the system monitors behaviors of the user in the recommendation list, low-intention behaviors and high-intention behaviors are distinguished, the high-intention behaviors trigger interest label weight updating, meanwhile, attenuation processing is executed on interest labels without behaviors for a long time, and a weight capping value is set; when the user actively cancels a single tag or all tags, the system stops tag recommendation or switches to first-level classification bottom recommendation; the updated interest label set and the corresponding weight participate in the next round of recommendation, dynamic closed-loop recommendation is achieved, and it is ensured that the recommendation result continuously fits the user interest and behavior data.

Description

Technical Field

[0001] This invention relates to the field of information recommendation technology, specifically to a user interest modeling method and system based on user explicit interest tags and high-intent behavior weighting, for achieving efficient matching and accurate recommendation between users and information. Background Technology

[0002] Most existing recommendation systems make recommendations based on users' historical behavior data or interest tags automatically generated by the system, but the following problems still exist: 1. Uncontrollable interest tags: Users cannot choose or adjust their areas of interest, and the interest tags predicted by the system may deviate from the user's true intentions; 2. Inaccurate weighting strategy: Most recommendation systems update interest weights when users only perform low-intent behaviors such as browsing or clicking, which can easily lead to recommendation results that deviate from the user's actual needs; 3. Lack of dynamic adjustment mechanism: When users' interests change or they actively unfollow content, the system cannot adjust the recommendation strategy in a timely manner, which may result in information redundancy or incorrect recommendations; Therefore, existing technologies cannot simultaneously achieve user autonomy in controlling their interests, dynamic adjustment of weights, and accurate matching and recommendation based on high-intent behavior triggers. Summary of the Invention

[0003] Technical solution To address the above problems, this invention provides a user interest modeling method and system based on explicit user interest tags and high-intent behavior weighting, the technical solution of which includes the following steps: 1. Users can choose their own interest tags. Users can add or remove interest tags on the information card page or the personalization settings page; 2. The system matches and filters platform information based on the user's set of interest tags to form a candidate information set; 3. Sort the candidate information set, generate a recommendation list, and display it to the user; 4. The system monitors user interaction behavior in the recommendation list and updates the weight of the corresponding interest tags only when the user generates high-intent interaction behavior; 5. Implement dynamic control over the weight of interest tags, including stopping recommendations when a tag is removed, weight decay for ineffective behavior, and weight cap control; 6. The updated interest tag weights participate in the next round of recommendation calculations, forming a dynamic closed-loop recommendation mechanism. Attached Figure Description

[0004] Figure 1This is a schematic diagram of the overall process of a user interest modeling method and system based on explicit user interest tags and high-intent behavior weighting according to the present invention. like Figure 1 As shown, the method includes the following steps: Step S1: The user enters the recommendation page; Step S2: Obtain the user's explicitly selected interest tags and read the corresponding weights; Step S3: Information matching and retrieval based on user interest tags; Step S4: Calculate the ranking score for the matched candidate information set and generate a recommendation list; Step S5: Monitor user behavior in the recommendation list; Step S6: Update the weights of the corresponding interest tags based on the user's high-intent behavior; Step S7: Use the updated set of interest tags to perform the next round of recommendations, thus achieving a dynamic closed loop; The above process also includes the following optional control branches: When a user actively cancels an interest tag, the system stops making recommendations based on that interest tag; When a user cancels all interest tags, the system makes recommendations based solely on primary category information. Figure 2 This is a schematic diagram of the process for obtaining and maintaining user-selected interest tags in this invention; like Figure 2 As shown, users can add or remove interest tags through the information card page or the personalization settings page; the system receives the user's confirmed interest tag operation results and stores and maintains the current interest tag set and its corresponding weights for subsequent information matching, sorting calculation and recommendation generation; Figure 3 This is a schematic diagram illustrating the process of updating the weight of interest tags based on user behavior in this invention. like Figure 3 As shown, the system monitors user interaction behavior in the recommendation list and distinguishes between low-intent behavior and high-intent behavior. Only when high-intent behavior is detected does the system update the weight of the corresponding interest tag. At the same time, the system performs weight decay processing on interest tags that have not generated effective behavior for a long time, and stops weight updates and recommendation calculations based on interest tags when users actively cancel interest tags. Detailed Implementation

[0005] The specific embodiments of the present invention will be described below with reference to the accompanying drawings: S1: User enters the recommendation page When a user enters the platform's homepage or information list page, the system reads the user's basic information and current set of interest tags, and determines whether the user is a new user or a low-behavior user, in order to initialize the weight of the interest tags. S2: Obtain user-selected interest tags and read their weights. Users can add or remove interest tags through the information card page or the personalization settings page; The system receives the user's confirmed interest tag operation results, uses the current set of interest tags and their corresponding weights for subsequent recommendation calculations, reads the historical weights of existing interest tags, and sets the initial weights for newly added interest tags; S3: Tag Matching Recall The system filters information from the platform's information set that matches at least one interest tag in the user's interest tag set to form a candidate information set. This step only performs matching recall and does not sort the information. S4: Calculate the ranking score and generate a recommendation list. The system calculates a ranking score for candidate information sets based on factors such as user interest tag weights, tag matching degree, freshness, and content quality, and generates a recommendation list according to the ranking score. During the ranking process, the system can be configured with exploration control logic and a minimum exposure mechanism for new content. S5: User Behavior Monitoring The system monitors user behavior in the recommendation list. When low-intent behaviors such as browsing are detected, the interest tag weights are not updated. When high-intent behaviors such as online chatting or exchanging contact information are detected, the weight update logic is triggered. S6: Update interest tag weights based on high-intent behavior The system only increases the weight of interest tags that actually match the information corresponding to high-intent behaviors; When a user actively cancels an interest tag, the system stops making recommendations and calculating weights based on that interest tag. For interest tags that have not generated effective behavior for a long period of time, the system performs weight decay processing; At the same time, an upper limit is set on the weight of interest tags to avoid the recommendation results becoming rigid due to the excessive weight of a single interest tag. S7: Moving on to the next round of recommendations Based on the updated set of user interest tags and their corresponding weights, the system re-executes matching and sorting to form a continuously updated closed-loop recommendation mechanism.

[0006] Implementation effect Through the above technical solution, the present invention can achieve the following effects: 1. Supports users' independent control over their interests, enabling the dynamic addition and removal of interest tags; 2. Update the weight of interest tags only when triggered by high-intent behavior to improve recommendation accuracy; 3. Dynamic adjustment of recommendation results is achieved through weight decay and upper limit control mechanisms; 4. Through a closed-loop feedback mechanism, the recommendation results continuously reflect the user's true interests.

Claims

1. A user interest modeling method and system based on user explicit interest tags and high intention behavior weighting, characterized in that, Includes the following steps: S1, the user enters the recommendation page; S2, obtain the set of interest tags selected and confirmed by the user, and read the corresponding tag weights; S3, Based on the set of interest tags, match and recall platform information to generate a set of candidate information; S4, sort the candidate information set to generate a recommendation list; S5 monitors user interaction behavior in the recommendation list and distinguishes between low-intent behavior and high-intent behavior. S6, update the interest tag weight corresponding to the behavior only when a user behavior that meets the preset high intent interaction conditions is detected; S7 executes the next round of recommendations based on the updated set of interest tags and their corresponding weights, forming a dynamic closed-loop recommendation.

2. The method according to claim 1, characterized in that, The interest tags are added or removed by the user through the information card page or the personalization settings page, and take effect after the user confirms; The system incorporates the user's currently confirmed interest tags and their corresponding weights into the user's tag set. For existing interest tags, the system reads the historical weights, and for newly added interest tags, it sets the initial weights.

3. The method according to claim 1, characterized in that, The high-intent interaction conditions include, but are not limited to, users initiating online chats, exchanging contact information, or other interactive behaviors that indicate that users have a clear intention to matchmake.

4. The method according to any one of claims 1 to 3, characterized in that, For interest tags that do not trigger high-intent interaction behavior within a preset time period, the system performs a weight decay process on their corresponding weights and sets an upper limit on the weights of interest tags to prevent the recommendation results from becoming rigid due to excessively high weights of a single interest tag.

5. The method according to any one of claims 1 to 4, characterized in that, When a user actively cancels an interest tag, the system stops information matching, recommendation ranking, and weight updates based on that interest tag.

6. The method according to any one of claims 1 to 5, characterized in that, When a user cancels all interest tags, the system implements a fallback recommendation strategy to ensure the integrity and basic diversity of the recommendation list.

7. An explicit interest tag based user precision matchmaking recommendation system characterized in that, include: The tag management module is used to receive interest tags selected and confirmed by users, and to maintain the user's interest tag set and corresponding weights. The information matching module is used to match and recall platform information based on the user interest tag set, and generate a candidate information set; The sorting and recommendation module is used to sort the candidate information set, generate a recommendation list, and present it to the user. The behavior monitoring module is used to monitor the user's interaction behavior in the recommendation list and distinguish between low-intent behavior and high-intent behavior. The weight update module is used to update the weight of the interest tag corresponding to the user behavior only when a user behavior that meets the preset high intent interaction conditions is detected. The recommendation control module is used to execute the next round of recommendations based on the updated set of interest tags and their corresponding weights, forming a dynamic closed-loop recommendation.

8. The system according to claim 7, characterized in that, The tag management module is configured as follows: When a user actively cancels an interest tag, that interest tag will no longer participate in subsequent information matching, recommendation ranking, and weight updates.

9. The system according to claim 7, characterized in that, The high-intent interaction conditions include, but are not limited to, users initiating online chat, exchanging contact information, submitting matchmaking requests, or other interactive behaviors that indicate that users have a clear intention to match.

10. A computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the method of any one of claims 1 to 6.

Images