Text Classification via Knowledge Base Filtering and Reasoning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing text classification methods face inefficiencies due to overwhelming amounts of redundant data in knowledge bases, which can overwhelm transformers and require extensive training data to accurately match text with relevant information, leading to suboptimal classification performance.
Innovation Solution
A method that filters a textual knowledge base to extract a subset of relevant facts and rules based on a textual query, using forward and backward reasoning to generate new facts and simplify queries, thereby improving classification accuracy through logical inference.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the entire textual knowledge base is used for classification, then more comprehensive information is available, but the transformer becomes overwhelmed and training efficiency decreases
Solution Approach 1:
The patent segments the large textual knowledge base into smaller, relevant subsets by filtering out irrelevant information before processing. This segmentation reduces the volume of data fed to the transformer while maintaining the essential information needed for accurate classification, thereby reducing computational burden without sacrificing classification reliability.
Solution Approach 2:
The patent extracts and removes irrelevant information from the textual knowledge base through filtering mechanisms. By taking out unnecessary data points, the system prevents the transformer from being overwhelmed by redundant information, allowing it to focus computational resources on processing only the relevant knowledge needed for accurate classification.
2Reliability
If extensive training data is used to match text with relevant information, then classification accuracy may improve, but training time and computational resources increase
Solution Approach 1:
The patent performs preliminary filtering of the textual knowledge base before the classification process begins. By pre-processing the knowledge base to extract only relevant information, the system reduces the amount of training data needed, thereby decreasing training time and computational resource requirements while maintaining or improving classification accuracy.
Solution Approach 2:
The patent extracts only the relevant portions of the textual knowledge base that are necessary for classification tasks. This extraction process eliminates redundant training data, reducing the time and computational resources required for training while preserving the essential information needed for accurate text matching and classification.
3Quantity of substance
If redundant data is included in the knowledge base, then more potential information is available, but classification performance becomes suboptimal
Solution Approach 1:
The patent systematically extracts and removes redundant information from the textual knowledge base through filtering processes. By eliminating duplicate and irrelevant data while preserving essential knowledge, the system improves classification performance by providing the transformer with clean, concentrated information that directly supports accurate text classification.
Solution Approach 2:
The patent segments the knowledge base to separate relevant from irrelevant information through filtering. This segmentation process divides the large body of knowledge into useful and redundant portions, allowing the system to process only the valuable segments that contribute to optimal classification performance while discarding the rest.
Data Source
AI summary
Methods, computer program products, and/or systems are provided that perform the following operations: obtaining a textual knowledge base; filtering the textual knowledge base to obtain a subset of the textual knowledge base, wherein the filtering is based on textual query data; generating reasoning data based on the subset of the textual knowledge base and the textual query data; generating classification data based on the subset of the textual knowledge base, the textual query data, and the reasoning data; and providing label data as output for the textual query data based on the classification data.