Content Recommendation Using Contextual LSA
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing content recommendation systems based on textual content descriptions suffer from low accuracy, particularly for video content, as they fail to properly link related or thematic content, especially when text descriptions are short and lack sufficient information.
Innovation Solution
A content recommendation method that incorporates both content text documents and external context documents, such as newspaper articles, to generate a unified mathematical representation using Latent Semantic Analysis (LSA) and Singular Value Decomposition (SVD), enhancing the identification of semantic relations and improving recommendation accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If recommendation systems use only content text documents, then the system is simple and fast, but the accuracy of content linking and recommendation relevance is low
Solution Approach 1:
The patent merges content text documents with external context documents (newspaper articles) into a unified mathematical representation using Latent Semantic Analysis. This combination allows the system to capture semantic relations between content items through shared contextual information, thereby improving recommendation accuracy without requiring a completely new system architecture.
Solution Approach 2:
The patent introduces external context documents as intermediary elements that mediate between content text documents. These context documents serve as a bridge to establish semantic connections between content items that would not be directly apparent from their own text descriptions, enhancing the recommendation system's ability to link related content.
2Reliability
If the system processes only content text documents, then processing time is short, but semantic relations between content are not properly identified
Solution Approach 1:
The patent performs preliminary processing by pre-processing both content text documents and external context documents into a unified mathematical representation before actual recommendation queries are processed. This pre-computation of semantic relationships allows for faster query responses while maintaining high reliability in semantic relation identification.
Solution Approach 2:
The unified mathematical representation created by combining content and context documents serves multiple functions: it enables semantic relation identification, supports recommendation generation, and allows for efficient querying. This multi-functionality reduces the need for separate processing steps, thereby maintaining processing time efficiency while improving semantic relation identification reliability.
Data Source
AI summary
A method of content recommendation, includes: generating a first digital mathematical representation of contents to associate the contents with a first plurality of words describing the contents; generating a second digital mathematical representation of text documents different from the contents to associate the documents with a second plurality of words; processing the first and second pluralities of words to determine a common plurality of words; processing the first and second digital mathematical representations to generate a common digital mathematical representation of the contents and the text documents based on the common plurality of words; and providing content recommendation by processing the common digital mathematical representation.