Automatic Ontology Generation via Embedding Representations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Novice users and sellers on online platforms face difficulties in effectively describing items, categorizing attributes, and setting prices, leading to challenges in finding buyers and matching desired items, which affects sales timing and buyer satisfaction.
Innovation Solution
The implementation of automatic ontology generation through embedding representations, which preprocesses item information into numerical formats, enabling machine-learning classifiers to predict attributes like brand and category, and named-entity recognition to tag relevant details, facilitating better item representation and search functionality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If automatic ontology generation is implemented, then item representation accuracy is improved, but system complexity increases
Solution Approach 1:
The patent introduces an automatic ontology generation system that acts as an intermediary between raw item data and the classification/search systems. This ontology layer preprocesses and structures item attributes, relationships, and hierarchies, thereby improving representation accuracy while shielding the core systems from direct complexity of raw data processing.
Solution Approach 2:
The system performs preliminary processing of item information by generating ontologies in advance. This includes pre-defining attribute structures, relationships, and hierarchies before actual classification or search operations occur, which improves downstream accuracy while containing complexity in a separate preprocessing stage.
2Ease of operation
If manual item description is used by novice users, then ease of operation is maintained, but item listing quality deteriorates
Solution Approach 1:
The patent enables novice users to list items effectively without manual expertise by implementing automatic ontology generation that self-structures item attributes. The system automatically categorizes, tags, and organizes item information based on learned ontologies, allowing users to simply provide raw item data while the system handles the complex structuring independently.
Solution Approach 2:
The patent replaces manual mechanical processes of item description and categorization with automated computational processes. Machine learning models and ontology generation algorithms substitute for human expertise in structuring item information, thereby maintaining ease of operation for users while dramatically improving listing quality through automated attribute extraction and relationship modeling.
3Ease of operation
If text searching is used by buyers, then ease of operation is maintained, but search accuracy deteriorates
Solution Approach 1:
The patent introduces an ontology-based intermediary layer between user search queries and item databases. This ontology layer provides structured semantic relationships and attribute hierarchies that enhance text search accuracy without requiring users to change their simple text-based search interface, thereby maintaining ease of operation while improving match precision through semantic understanding.
Data Source
AI summary
Disclosed herein are system, computer-readable storage medium, and method embodiments of automatic ontology generation by embedding representations. A system including at least one processor may be configured to receive a vectorized feature set derived from an embedding and including first and second features, and provide the vectorized feature set to a fuser set including first and second fusers. The system may be configured to generate a representation from the fuser set based on the first and second features, and derive tasks based on the representation, assigning to the tasks respective qualifier sets including a weight value, a loss function, and a feedforward function. The system may be configured to compute respective weighted losses for the tasks, based on the respective qualifier sets, and output a data model based on backpropagating the respective weighted losses through the fuser set, the vectorized feature set, the embedding, or a combination thereof.