Media Content Representation Using Heterogeneous Query Graphs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing media content recommendation models rely on strict word-by-word matching, leading to improper content recommendations when similar words with consistent intentions are inputted, limiting the recommendation process.
Innovation Solution
Construct heterogeneous graphs using meta-paths to represent user and media content information, determining and storing representation information to build a media content feature library, enabling more accurate and efficient media content recommendations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If text retrieval manner is used for media content recommendation, then the recommendation process is simple and fast, but the recommendation accuracy deteriorates when similar words with consistent intentions are inputted
Solution Approach 1:
The patent transforms the recommendation approach from traditional text retrieval based on exact word matching to semantic representation learning where media content and query information are converted into vector representations. This parameter change allows the system to capture semantic meanings rather than relying on exact word matches, thereby improving recommendation accuracy while maintaining computational efficiency through vector similarity calculations.
Solution Approach 2:
The patent introduces heterogeneous graphs as an intermediary structure that connects query information and media content through multiple meta-paths. These graphs serve as mediators that capture complex relationships and semantic associations between queries and media content, enabling the system to retrieve relevant content even when exact word matches are absent, thus resolving the contradiction between simple retrieval and accurate recommendation.
2Measurement precision
If heterogeneous graphs with multiple meta-paths are constructed to improve recommendation accuracy, then the recommendation accuracy is improved, but the system complexity increases
Solution Approach 1:
The patent segments the complex heterogeneous graph construction into multiple distinct meta-paths, each representing specific semantic relationships (e.g., query-media, query-entity, entity-media). This segmentation allows the system to build comprehensive semantic representations through structured, modular components rather than attempting to process all relationships simultaneously, thereby managing system complexity while maintaining high recommendation accuracy.
Solution Approach 2:
The patent creates a universal heterogeneous graph framework that can handle multiple types of relationships and data sources through a unified meta-path structure. This multi-functional approach allows the same graph construction methodology to process diverse information types (queries, media content, entities, attributes) consistently, reducing overall system complexity by providing a single versatile solution rather than separate processing mechanisms for each data type.
Data Source
AI summary
A media content processing method includes obtaining a first heterogeneous graph and a second heterogeneous graph. Obtaining the first heterogeneous graph includes constructing a first meta-path. The first meta-path includes a first start information node representing historical query information and at least one first other information node representing media information. The media information includes one media content and associated information about the media content. The media content is in a query result corresponding to the historical query information and is operated by a user. Obtaining the second heterogeneous graph includes constructing a second meta-path. The second meta-path includes a second start information node representing the media content and at least one second other information node representing the query information or the associated information. The method further includes determining and storing, based on the first heterogeneous graph and the second heterogeneous graph, representation information corresponding to the media content.