Recommender System Metadata Transformation for Cross-Domain Profiles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing recommender systems face challenges in providing consistent and high-quality content recommendations across different domains, such as live TV and video on demand, due to isolated user profiles and weak cross-domain recommendation performance.
Innovation Solution
The solution involves transforming and enriching metadata to ensure user profiles are applicable across domains, using techniques like Extensible Stylesheet Language Transformation (XSLT) to standardize metadata formats, and applying ontologies for classification enrichment, allowing for seamless recommendations across various content sources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If isolated user profiles are used for each domain (live TV, VoD, internet), then domain-specific recommendation accuracy is improved, but cross-domain recommendation performance deteriorates
Solution Approach 1:
The patent merges isolated domain-specific user profiles into a unified cross-domain user profile by transforming and enriching metadata from multiple domains (live TV, VoD, internet) into a common format. This allows the recommender engine to leverage user preferences across all domains while maintaining domain-specific characteristics, thereby improving cross-domain recommendation performance without sacrificing domain-specific accuracy.
Solution Approach 2:
The patent creates a universal metadata transformation framework that enables a single recommender engine to function across multiple domains. By defining domain-independent metadata attributes and transformation rules, the system achieves multi-functionality where the same recommendation infrastructure serves live TV, VoD, and internet content domains effectively.
2Reliability
If domain-specific metadata formats are maintained, then domain-specific content characteristics are preserved, but recommender engine applicability across domains deteriorates
Solution Approach 1:
The patent introduces metadata transformation and enrichment as an intermediary layer between domain-specific content sources and the universal recommender engine. This intermediary process transforms diverse metadata formats (EPG, VoD, internet) into a standardized common format while preserving essential domain-specific characteristics through carefully selected domain-independent attributes, enabling seamless recommender engine applicability across domains.
Solution Approach 2:
The patent applies parameter transformation by converting domain-specific metadata parameters into domain-independent parameters. The transformation process maps various source formats to a unified schema with standardized attributes, allowing the recommender engine to operate consistently across different content domains while maintaining the essential characteristics of each domain through appropriate parameter selection and weighting.
3Measurement precision
If multiple isolated recommender systems are deployed for each domain, then domain-specific recommendation quality is improved, but system complexity deteriorates
Solution Approach 1:
The patent implements a universal recommender engine that serves multiple domains through a single unified system. By establishing a common metadata transformation framework and domain-independent attribute schema, the system eliminates the need for separate recommender deployments in each domain, thereby reducing system complexity while maintaining domain-specific recommendation quality through the transformation layer.
Solution Approach 2:
The patent segments the recommendation system into distinct functional layers: domain-specific metadata collection, transformation/enrichment layer, and universal recommender engine. This segmentation allows each layer to be optimized independently while maintaining overall system efficiency, reducing complexity by separating domain-specific concerns from the core recommendation logic.
Data Source
AI summary
The present invention relates to an apparatus, a method and a computer program product for controlling distribution and processing of content-related metadata of different content sources (101-1 to 101-n), wherein extracted metadata are transformed at from a received format of the extracted metadata into a predetermined common format used by a recommender engine. Additionally or alternatively, it is checked for an overlap between a detected classification parameter and at least one stored classification parameter, and non-overlapping metadata is enriched by adding at least one new classification parameter derived from an ontology-based processing of the non-overlapping metadata.