Question Answering System Using Text Entailment Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing question answering systems overwhelm users with excessive information and fail to reflect the meaning of natural language inputs, requiring users to sift through keyword-based search results to find relevant answers.
Innovation Solution
A question answering system that generates a candidate answer list by performing text entailment recognition on natural language questions, combining evidence sentences based on association degrees, and providing a user-friendly output that ranks answers by relevance, using word, sentence, and context entailment recognition techniques.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If keyword-based search method is used, then search coverage is improved, but information overload increases and user fatigue increases
Solution Approach 1:
The patent extracts only the most relevant information from search results by using question-answer pair matching and relevance scoring. Instead of displaying all keyword-matched documents, the system extracts and prioritizes answers that directly address the user's question, filtering out excessive information while maintaining comprehensive search coverage.
Solution Approach 2:
The patent introduces an intermediary processing layer between keyword search and result display. This layer includes question-answer pair extraction, relevance scoring mechanisms, and ranked listing that mediates between comprehensive search results and user-friendly presentation, reducing information overload while preserving search versatility.
2Device complexity
If keyword extraction method is used, then search processing is simplified, but natural language meaning is not reflected
Solution Approach 1:
The patent changes the parameter of text processing from simple keyword extraction to question-answer pair extraction with semantic analysis. By transforming the processing parameter from keyword-based to meaning-based pairing, the system preserves natural language meaning while maintaining manageable processing complexity through structured extraction rules.
Solution Approach 2:
The patent performs preliminary extraction of question-answer pairs from documents before the search query is processed. By pre-identifying and structuring question-answer relationships in the document corpus, the system enables meaning-preserving search without adding complexity during the actual query processing phase.
3Quantity of substance
If ranked documents list is provided, then search completeness is improved, but user effort to find answer increases
Solution Approach 1:
The system performs preliminary ranking and selection of answer candidates before presenting results to users. By pre-computing relevance scores and identifying the most likely answers from the complete search results, the system reduces user search time while maintaining result completeness through ranked presentation.
Solution Approach 2:
The patent segments the complete search results into ranked answer candidates with relevance indicators. Instead of presenting a flat list of documents, the system segments results by relevance score and answer quality, allowing users to quickly identify and access the most relevant answers while preserving access to complete results if needed.
Data Source
AI summary
Provided is a question answering system with respect to a natural language question and a method thereof. The question answering system includes a candidate answer generating unit configured to extract a document mapped to an input natural language question, and generate candidate answers with respect to the natural language question from the extracted document, a text entailment recognizing unit configured to generate a text entailment recognition result representing a degree of association between multiple evidence sentences including the generated candidate answers and the natural language question, a list generating unit configured to generate a candidate answer list including the multiple evidence sentences in high association degree order on the basis of the text entailment recognition result, and an output unit configured to output the generated candidate answer list as a search result with respect to the natural language question.