Information Extraction System for Problem-Evoking Expression Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing information extraction methods are inadequate for identifying problem-evoking expressions that superficially match dictionary entries but vary in representation based on context or background knowledge, leading to incorrect classification of positive and negative examples.
Innovation Solution
An information extraction system that acquires and compares sentence sets matching positive and negative example patterns from a corpus to specify constituent elements characterizing problem-evoking expressions, distinguishing between problematic and non-problematic instances.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If dictionary-based superficial matching is used to identify problem evoking expressions, then the extraction process is simple and fast, but the classification precision deteriorates because many words are incorrectly assigned false classes
Solution Approach 1:
The patent introduces solution request sentences as an intermediary to bridge the gap between superficial dictionary matching and accurate problem identification. These sentences serve as mediators that connect the expression to its actual problem context, enabling precise classification without sacrificing extraction efficiency
Solution Approach 2:
The system performs preliminary extraction of solution request sentences containing problem evoking expressions before final classification. This preliminary action separates potentially problematic expressions from benign ones early in the process, allowing subsequent precise classification based on contextual analysis
2Productivity
If clustering is performed on training data candidates including false data, then the processing is efficient, but the identification information quality deteriorates because false classes dominate the clusters
Solution Approach 1:
The patent extracts and removes false data from training data candidates before performing clustering. By taking out incorrectly assigned examples, the system ensures that clusters are formed only from reliable positive and negative examples, maintaining both processing efficiency and identification quality
Solution Approach 2:
The system performs preliminary filtering to remove false data candidates before the clustering process. This preliminary action prevents false classes from dominating clusters, ensuring that clustering operates on high-quality data while maintaining processing efficiency
3Device complexity
If single search purpose is used to identify positive examples, then the search is simple, but the reliability deteriorates because opposite requests cannot be distinguished
Solution Approach 1:
The patent segments the search process into multiple purposes: identifying positive examples containing problem evoking expressions and identifying negative examples containing opposite requests. This segmentation allows the system to distinguish between problematic and non-problematic instances reliably while maintaining manageable search complexity through systematic organization
Data Source
AI summary
A system, method and program are provided. For example, a method includes acquiring sentence sets, respectively matching a positive or negative example solution request pattern which represents a positive or negative example of a sentence including a problem evoking expression, the negative example being an opposite request to the positive example, from a corpus respectively as a positive and negative example solution request sentence sets, extracting, respectively, the problem evoking expression from a sentence in the positive example solution request sentence set, and a sentence in the negative example solution request sentence set. The method further includes comparing, for each problem evoking expression, constituent elements of sentences included in the positive and negative example solution request sentence sets, and specifying respective constituent elements characterizing the positive or negative example solution request sentence set respectively as positive or negative example identification information.