Lexical Answer Type Identification for Conditional Question Answering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current question answer systems face difficulties in interpreting questions with conditions, leading to incorrect answers due to text similarity between questions and document passages.
Innovation Solution
A knowledge manager system that identifies lexical answer types in questions, applies condition annotations, and generates condition fulfillment scores to rank candidate answers based on their relevance to the conditions within the question.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If current question answer systems rely on text similarity between questions and document passages, then they can provide answers based on surface-level matching, but they produce incorrect answers when questions contain conditions that are not explicitly matched
Solution Approach 1:
The patent segments the question analysis process into distinct components: identifying lexical answer types, detecting conditions, generating candidate answers, and evaluating condition fulfillment. This segmentation allows the system to handle complex conditional questions by breaking them down into manageable processing stages, improving answer accuracy without overwhelming system complexity
Solution Approach 2:
The patent introduces condition annotations as an intermediary layer between the question and candidate answers. These annotations serve as mediators that carry condition information through the system, enabling the evaluation of whether candidate answers satisfy the conditions specified in the question, thereby improving answer precision
2Measurement precision
If the system evaluates candidate answers against multiple conditions with condition fulfillment scores, then answer precision improves, but processing time and computational resources increase
Solution Approach 1:
The patent implements partial evaluation by generating condition fulfillment scores for candidate answers based on the degree to which they satisfy conditions. Rather than requiring absolute proof of condition satisfaction, the system uses scoring to rank candidates, which reduces processing time while maintaining sufficient precision for practical applications
Solution Approach 2:
The system performs preliminary identification of lexical answer types and conditions before generating candidate answers. This preliminary action allows the system to filter and prepare evaluation criteria in advance, reducing the computational burden during the actual answer evaluation and ranking phase
Data Source
AI summary
An approach is provided in which a knowledge manager receives a question and selects candidate answers based upon a lexical answer type included in the question. The knowledge manager evaluates the candidate answers against conditions detected in the question and generates condition fulfillment scores accordingly. In turn, the knowledge manager ranks the candidate answers based upon the corresponding condition fulfillment scores.