Question Answering System Using Hypothesis Pruning for Computational Cost
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current question answering systems face challenges in understanding natural language queries and providing accurate answers from large collections of structured and unstructured documents, as they lack human-like natural language processing and common sense reasoning, leading to inefficiencies in identifying relevant information.
Innovation Solution
A method and system that generate and evaluate candidate answers by filtering and ranking them based on preliminary scores, using supporting passage retrieval and logistic regression models to optimize computational cost and answer quality, focusing on high-quality candidate answers for further investigation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all candidate answers are processed to retrieve supporting evidence and compute final scores, then answer quality and reliability improve, but computational cost and processing time increase significantly
Solution Approach 1:
The patent applies preliminary action by computing preliminary scores for all candidate answers before retrieving supporting evidence. This early scoring step filters out low-quality candidates that are unlikely to be correct, so that full processing (evidence retrieval and final scoring) is only applied to a small subset of high-potential candidates. This resolves the contradiction by performing preliminary assessment to avoid wasting computational resources on poor candidates while maintaining reliability through thorough processing of promising candidates.
Solution Approach 2:
The patent implements partial action by processing only a subset of candidate answers with full rigor (supporting evidence retrieval and final scoring), while applying a lighter preliminary screening to all candidates. The system performs exactly the amount of processing needed - not more, not less - by using preliminary scores to identify which candidates warrant full processing. This optimizes the balance between answer quality and computational cost.
2Difficulty of detecting and measuring
If comprehensive natural language processing and common sense reasoning are implemented, then understanding of queries improves, but system complexity increases
Solution Approach 1:
The patent applies segmentation by breaking down the query processing into distinct modular components: query analysis, candidate answer generation, preliminary scoring, supporting evidence retrieval, and final answer selection. Each module performs a specific function and can be independently optimized or replaced. This modular architecture reduces system complexity while maintaining comprehensive natural language processing capabilities, as each segment handles a specific aspect of understanding without requiring the entire system to be overly complex.
Data Source
AI summary
A method, system and computer program product for generating answers to questions. In one embodiment, the method comprises receiving a query, conducting a search through one or more data sources to identify candidate answers to the query, and providing each of the candidate answers with a preliminary score. The method further comprises filtering out any of the candidate answers with a preliminary score that does not satisfy a defined condition. The candidate answers having preliminary scores that satisfy this condition form a subset of the candidate answers. Each of the candidate answers in this subset is processed to produce further scores. A ranking function is applied to these further scores to determine a ranking for each of the candidate answers in the subset; and after this ranking function is applied, one or more of the candidate answers are selected as one or more final answers to the query.