Entity-Relation Passage Scoring for Question Answering Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current question answering computer systems face challenges in accurately determining the relevance of passages to queries due to their inability to effectively consider contextual and semantic relationships between entities and relations, leading to inefficient passage scoring.
Innovation Solution
The method involves parsing queries and passages to identify noun-centric and verb-centric phrases, creating entity and relation pairs, and determining a passage score based on matching criteria, using a combination of syntactic and semantic knowledge to align untyped entities and relations, thereby enhancing context-dependent scoring.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If keyword frequency-based scoring is used for searching documents, then the search process is simple and fast, but the system fails to consider the context of particular words and produces inaccurate results
Solution Approach 1:
The patent segments the passage scoring process into multiple independent components: entity identification, relation identification, and scoring computation. Each component processes specific aspects of the text separately before integrating results, allowing the system to maintain speed while improving accuracy through specialized processing of contextual information
Solution Approach 2:
The patent introduces an intermediary scoring function that mediates between the raw text and the final relevance score. This scoring function processes entity and relation pairs as intermediate representations, transforming them into meaningful relevance scores that capture contextual relationships without requiring complex direct analysis of the entire passage
2Measurement precision
If NLP techniques are used to analyze syntax and context, then the system can infer meaning and improve answer accuracy, but the complexity of human language makes the system slow and computationally intensive
Solution Approach 1:
The patent extracts only the most critical elements for meaning inference: entities and their relations. By focusing on these key components rather than analyzing every syntactic detail of the entire passage, the system achieves accurate meaning inference while significantly reducing processing complexity and computational requirements
Solution Approach 2:
The patent applies partial action by processing only the necessary portions of the text (entities and relations) rather than the entire passage. This selective processing approach maintains sufficient accuracy for meaning inference while avoiding the excessive computational cost of analyzing all syntactic and contextual details
3Adaptability or versatility
If multiple NLP techniques are combined to handle language complexity, then the system can better understand polysemous words and contextual nuances, but the system becomes slower and more complex
Solution Approach 1:
The patent merges multiple NLP techniques (entity identification, relation identification, and scoring) into a unified integrated system. This combination allows the system to handle polysemous words and contextual nuances through coordinated processing of entities and relations, achieving versatile language understanding while maintaining processing efficiency through the unified architecture
4Reliability
If the system processes all retrieved passages for evidence generation, then the completeness of answer evidence is high, but the time required for passage scoring increases significantly
Solution Approach 1:
The patent applies preliminary action by pre-processing passages to identify and extract entities and relations before the main scoring process. This preliminary extraction of meaningful components allows the system to quickly assess passage relevance without re-analyzing the entire passage text, thereby maintaining evidence completeness while significantly reducing scoring time
Data Source
AI summary
According to an aspect, a query and a passage are parsed by a language parser to detect noun-centric phrases and verb-centric phrases in the query and the passage. Entities, including at least one untyped entity, are identified based on the noun-centric phrases and relations are identified based on the verb-centric phrases. Entity pairs are created that include an entity identified in the query and an entity identified in the passage, each pair satisfies a matching criteria. Relation pairs are created that include a relation identified in the query and a relation identified in the passage, each pair satisfies a matching criteria. A passage score that indicates the likelihood that an answer to the query is contained in the passage is determined based on the entity pairs, the matching criteria satisfied by each entity pair, the elation pairs, and the matching criteria satisfied by each relation pair.