Automatic Taxonomy Mapping via Sequence Semantic Embedding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for mapping taxonomies across different systems are inefficient and require manual human intervention, leading to disparities and outdated mappings due to variations in terminology and frequent updates, which hampers the ability of publication sites to provide accurate recommendations to users.
Innovation Solution
The implementation of sequence semantic embedding (SSE) technology, which automates the mapping of taxonomies by projecting inventory taxonomy entries into a shared semantic vector space, allowing for automated comparison and updating of taxonomies across different systems, eliminating the need for manual mapping and enabling scalability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual mapping methods with lexical level matching and regular expression rules are used, then mapping can be performed with existing systems, but the mapping becomes outdated quickly and requires continuous manual updates due to terminology variations and frequent taxonomy updates
Solution Approach 1:
The system performs self-service taxonomy mapping by automatically comparing taxonomy entries across systems using sequence semantic embedding. The neural network model autonomously identifies corresponding categories without human intervention, updating mappings dynamically as taxonomies change, thereby eliminating the need for continuous manual updates while maintaining high accuracy.
Solution Approach 2:
The patent replaces manual mechanical mapping processes with an automated neural network-based semantic embedding system. Instead of human operators manually creating and updating mapping rules, the system uses deep learning models to automatically compute semantic similarities and generate mappings, substituting human labor with intelligent automation.
2Ease of manufacture
If lexical level matching with manually crafted mapping files and regular expression rules is used, then mapping can be established between taxonomies, but the approach cannot be shared across different content generators due to terminology variations
Solution Approach 1:
The sequence semantic embedding model provides a universal mapping mechanism that works across different content generators and taxonomy systems. By converting taxonomy entries into semantic vectors that capture their meaning rather than relying on specific lexical patterns, the system adapts to various terminology styles and structures without requiring generator-specific customization, thereby achieving broad applicability.
Solution Approach 2:
The system changes the parameter of comparison from lexical similarity (surface-level word matching) to semantic similarity (meaning-based vector comparison). This parameter transformation allows the mapping approach to remain effective across different content generators with varying terminology, as the semantic embedding captures the underlying meaning independent of specific word choices.
3Measurement precision
If human manual involvement is used for taxonomy mapping, then accurate mappings can be created initially, but the process becomes inefficient and cannot keep pace with frequent taxonomy updates
Solution Approach 1:
The patent substitutes manual human mapping operations with an automated neural network system that performs semantic embedding and comparison. This automation maintains high mapping accuracy through sophisticated semantic analysis while dramatically increasing productivity by processing taxonomy updates instantly without human intervention, keeping pace with frequent changes.
Solution Approach 2:
The system enables continuous automatic updating of taxonomy mappings as taxonomies evolve. Rather than periodic manual updates, the neural network continuously processes new taxonomy versions, maintaining accurate mappings in real-time. This continuous automated action ensures mappings stay current with frequent taxonomy changes while preserving high accuracy through consistent semantic analysis.
Data Source
AI summary
In accordance with an example embodiment, an automated taxonomy mapping system that uses sequence semantic embedding techniques is described. Sequence sematic embedding models are used to generate the sequence vectors. The sequence semantic embedding models are trained offline and can be shared across different systems having different taxonomies and various versions of a category taxonomy.