Neural Taxonomy Expansion Using Embedding-Based Parent Placement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automatically placing a taxonomy candidate within an existing taxonomy is a time-consuming and labor-intensive process, similar to classifying biological organisms, which hinders the efficiency of online recommender services in recommending relevant items to users.
Innovation Solution
A neural taxonomy expander is trained using a curated taxonomic hierarchy to generate embedding vectors for taxonomy candidates, projecting them into a taxonomic hyperspace, and identifying the closest neighbors to determine the correct parent node within the taxonomy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual curation is used to place taxonomy candidates into existing taxonomy, then classification accuracy is improved, but time consumption and labor intensity increase significantly
Solution Approach 1:
The patent introduces an intermediary neural network system that acts as a mediator between taxonomy candidates and the existing taxonomy hierarchy. The neural network processes candidate items through embedding vectors and projection functions, automatically determining their appropriate positions in the taxonomy without requiring manual human curation for each item, thus resolving the contradiction between accuracy and time consumption
Solution Approach 2:
The patent replaces the mechanical manual curation process with an automated neural network-based system. Instead of human experts manually reviewing and placing each taxonomy candidate, the system uses machine learning models (including embedding generation, projection tensors, and nearest neighbor search) to automatically perform the classification task, significantly reducing time consumption while maintaining accuracy
2Reliability
If manual curation is used to place taxonomy candidates, then classification reliability is improved, but productivity decreases
Solution Approach 1:
The patent implements a self-service taxonomy expansion system where the neural network automatically processes taxonomy candidates without requiring continuous human intervention. The system generates embedding vectors, projects them into the taxonomy hyperspace, identifies nearest neighbors, and places candidates autonomously, enabling high-volume processing while maintaining reliable classification through the trained model
Solution Approach 2:
The patent transforms the classification task from a manual process to an automated one by changing the operational parameters - using neural network embeddings and projection tensors instead of human judgment. This parameter change enables the system to process numerous candidates simultaneously through vector operations, dramatically increasing productivity while maintaining reliability through consistent algorithmic application
Data Source
AI summary
Systems and methods for automatically placing a taxonomy candidate within an existing taxonomy are presented. More particularly, a neural taxonomy expander (a neural network model) is trained according to the existing, curated taxonomic hierarchy. Moreover, for each node in the taxonomic hierarchy, an embedding vector is generated. A taxonomy candidate is received, where the candidate is to be placed within the existing taxonomy. An embedding vector is generated for the candidate and projected by a projection function of the neural taxonomy expander into the taxonomic hyperspace. A set of closest neighbors to the projected embedding vector of the taxonomy candidate is identified and the closest neighbor of the set is assumed as the parent for the taxonomy candidate. The taxonomy candidate is added to the existing taxonomic hierarchy as a child to the identified parent node.