Content Recommendation Using Look-Alike Model Attention Mechanism
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional content recommendation systems often struggle with accuracy, as they primarily focus on user interests based on historical behavior and semantic content features, neglecting the user's need for discovering current hotspot events and high-quality content from a small audience.
Innovation Solution
A content recommendation method and apparatus that utilize a look-alike model with an attention mechanism to calculate similarities between a target user vector and seed user vectors corresponding to candidate recommendation content, thereby improving the accuracy of content recommendation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional recommendation algorithms are used to recommend content based on user historical behavior and semantic features, then the recommendation system can operate with simple processing logic, but the recommendation accuracy is insufficient
Solution Approach 1:
The patent introduces seed user vectors as an intermediary representation between the target user and candidate content. Instead of directly computing the complex interaction between user historical behavior and content semantic features, the system uses seed user vectors (derived from users who interacted with the content) as a mediator to bridge the gap, thereby improving recommendation accuracy while maintaining manageable processing complexity
Solution Approach 2:
The patent transforms the recommendation problem into a different dimensional space by using user vectors and seed user vectors. The recommendation process is reframed as calculating similarity between the target user vector and seed user vectors in a high-dimensional vector space, which simplifies the complex direct user-content interaction modeling while improving accuracy
2Adaptability or versatility
If the recommendation system only considers user interest based on historical behavior, then the system structure remains simple, but it cannot meet user needs for discovering hotspot events and high-quality content
Solution Approach 1:
The patent segments the user interest representation into multiple components: the target user vector representing the user's historical interests, and seed user vectors representing the interests of users who interacted with specific content. This segmentation allows the system to simultaneously capture both the user's personal history and the content's appeal to specific user groups, thereby meeting diverse user needs including discovering hotspots and finding high-quality content
Solution Approach 2:
The system performs preliminary action by pre-calculating and storing seed user vectors for candidate content based on historical interaction data. This preliminary processing of content characteristics into seed user vectors enables the system to quickly and accurately match them with target users during the recommendation process, improving both adaptability to user needs and recommendation accuracy
Data Source
AI summary
This application discloses a content recommendation method performed at a computer device and belongs to the field of artificial intelligence. The method includes: acquiring a target user vector of a target user; determining n groups of seed user vectors according to the target user vector, each group of seed user vectors corresponding to a respective piece of candidate recommendation content; invoking a look-alike model to calculate a similarity between the target user vector and each group of seed user vectors, the look-alike model being used for calculating a similarity between user vectors based on an attention mechanism; and determining, among the n pieces of candidate recommendation content, target content to be recommended to the target user according to the respective similarities of the corresponding n groups of seed user vectors. This application can resolve a problem of relatively low accuracy of a recommendation method in the related art.