Attribute and Rating Co-Extraction via Multi-Sub-Network Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for extracting information from user reviews, such as attributes and ratings, are inefficient and do not allow for simultaneous extraction of attribute and rating information, which is essential for building comprehensive knowledge graphs of entities.
Innovation Solution
A method and system utilizing a multi-sub-network model that determines semantic feature representations, attribute information, and rating information from text, enabling simultaneous extraction of attributes and ratings through separate sub-networks for feature, attribute, and rating processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional separate extraction methods are used for attributes and ratings, then the extraction process is simpler, but the extraction efficiency is low and cannot extract both attribute and rating information simultaneously
Solution Approach 1:
The patent combines separate attribute extraction and rating extraction models into a unified co-extraction model. The model integrates multiple sub-networks (attribute extraction sub-network, rating extraction sub-network, and feature representation sub-network) that work together to extract both attribute and rating information simultaneously from the same input text, thereby improving extraction efficiency while maintaining organized internal structure.
Solution Approach 2:
The unified co-extraction model is segmented into distinct sub-networks with specific functions: a feature representation sub-network for generating semantic features, an attribute extraction sub-network for identifying attributes, and a rating extraction sub-network for determining ratings. This segmentation allows each component to specialize in its task while working collaboratively, resolving the contradiction between complexity and efficiency.
2Loss of information
If conventional extraction methods are used, then the processing time is shorter for single attributes, but the overall information extraction completeness is insufficient
Solution Approach 1:
The model enables continuous extraction of both attribute and rating information in a single processing pass through the unified architecture. The feature representation generated by the first sub-network is continuously utilized by both the attribute extraction sub-network and the rating extraction sub-network, eliminating the need for separate processing steps and ensuring complete information extraction without redundant time loss.
3Measurement precision
If multi-sub-network model is used for co-extraction, then the extraction accuracy is improved, but the computational complexity increases
Solution Approach 1:
The feature representation sub-network serves multiple functions by generating semantic features that are utilized by both the attribute extraction sub-network and the rating extraction sub-network. This multi-functionality reduces redundant computations and optimizes resource utilization, allowing the model to achieve high extraction accuracy while managing computational complexity through shared feature processing.
Data Source
AI summary
Embodiments of the present disclosure relate to attribute and rating co-extraction. According to embodiments of the present disclosure, a method is proposed. The method comprises: determining, by a first sub-network of a model, a first feature representation based on a first token contained in a text, the first feature representation indicating semantic information of the first token in the text; determining, by a second sub-network of the model, first attribute information associated with the first token based on the first feature representation, the first attribute information indicating a first attribute involved in the text; and determining, by a third sub-network of the model, first rating information associated with the first token based on the first feature representation, the first rating information indicating a rating related to the first attribute.