Question Answering System for Multiple Correct Answers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing question answering systems struggle to accurately handle multi-correct answer questions, where multiple answers can satisfy the query, especially in open domain scenarios like Wikipedia, leading to reliability issues and performance degradation.
Innovation Solution
A question answering system and method that performs symbolic and deep learning-based searches, ranks paragraphs, and uses a multi-task model to verify and present multiple correct answers by determining the support or opposition of searched paragraphs, employing a linear combination of re-ranking and machine reading probabilities to identify and merge supporting paragraphs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing question answering systems use traditional single-answer extraction methods, then the system complexity remains low, but the reliability deteriorates when multiple correct answers exist
Solution Approach 1:
The system segments the answer extraction process into multiple independent modules: paragraph search module, paragraph ranking module, answer extraction module, and answer verification module. Each module handles a specific aspect of the multi-correct answer problem, allowing the system to maintain high reliability while managing complexity through functional decomposition.
Solution Approach 2:
The patent introduces intermediate components including a paragraph embedding model that transforms paragraphs into vector representations, and a multi-task model that serves as an intermediary verification layer. These intermediaries enable the system to evaluate multiple potential answers objectively before final selection, improving reliability without proportionally increasing overall system complexity.
2Reliability
If the system performs comprehensive paragraph search and verification to identify multiple correct answers, then the reliability improves, but the processing time increases
Solution Approach 1:
The system performs preliminary paragraph search and embedding before answer extraction. By pre-processing paragraphs into vector representations and ranking them by relevance before the actual answer extraction and verification stages, the system reduces the computational burden during time-critical processing phases while maintaining comprehensive verification for reliability.
Solution Approach 2:
The patent implements a multi-task model that performs multiple verification functions simultaneously (answer extraction, yes/no verification, and multi-answer detection). This partial action approach allows the system to focus computational resources on the most critical verification tasks rather than performing exhaustive analysis on all possible answer variations, balancing reliability with processing efficiency.
3Productivity
If the system extracts all potential correct answers without verification, then the productivity increases, but the measurement precision deteriorates
Solution Approach 1:
The system implements a feedback-based verification mechanism where the multi-task model evaluates extracted answers against the original query and context. The verification module provides feedback on answer validity, allowing the system to quickly filter incorrect answers while maintaining high productivity. This feedback loop ensures measurement precision without requiring exhaustive manual verification of each potential answer.
Solution Approach 2:
The patent replaces manual or rule-based answer verification with deep learning-based automated verification using the multi-task model. This substitution enables rapid, accurate determination of answer correctness by leveraging learned patterns from training data, significantly improving both productivity and measurement precision compared to traditional mechanical verification methods.
Data Source
AI summary
The present disclosure relates to a question answering system and method capable of inferring multiple correct answers. The question answering system capable of inferring multiple correct answers according to the present disclosure includes an input interface device configured to receive a query, a memory for storing a program that analyzes the query and searches for a paragraph with a high probability of including a correct answer through a document search, and a processor for executing the program, wherein the processor extracts a correct answer using a result of the searching for a paragraph with a high probability of including a correct answer, determines whether the extracted correct answer corresponds to multiple correct answers, and provides a correct answer extraction result.