Knowledge Graph Fusion for Cross-Document Reasoning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional Q&A systems are not suitable for cross-document answer collation and often fail to obtain accurate answers from multiple documents, as they rely on organizing large corpora and conventional reading comprehension without effectively utilizing knowledge graphs for cross-document reasoning.
Innovation Solution
A method that extracts contextual information from multiple documents using natural language processing, generates knowledge graphs for each document, and fuses linked entities across these graphs to create a larger knowledgebase, enabling more accurate answers to natural language queries through cross-document reasoning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional Q&A systems organize large corpora and use conventional reading comprehension, then they can answer factual questions, but they fail to provide accurate answers for cross-document questions
Solution Approach 1:
The patent merges multiple document-specific knowledge graphs into a single fused knowledge graph that spans across documents. This consolidation allows the system to capture cross-document relationships and provide accurate answers to cross-document questions while avoiding the complexity of managing separate corpora for each document.
Solution Approach 2:
The patent introduces a knowledge graph fusion module as an intermediary between individual document processing and final question answering. This intermediary component links entities across documents by identifying and merging overlapping or related entities, enabling accurate cross-document reasoning without requiring complex corpus organization.
2Adaptability or versatility
If knowledge graphs are generated for each document separately, then contextual information can be extracted from multiple documents, but the system cannot effectively reason across documents
Solution Approach 1:
The patent combines multiple separate knowledge graphs into a fused knowledge graph that preserves and extends entity links across documents. This merging process maintains contextual information from each document while establishing cross-document relationships, enabling the system to reason across documents without losing information.
Solution Approach 2:
The fused knowledge graph serves multiple functions simultaneously: it stores contextual information from individual documents, establishes cross-document entity relationships, and provides a unified structure for answering questions. This multi-functionality enables cross-document reasoning while preserving all necessary contextual details.
3Measurement precision
If conventional reading comprehension is used, then factual questions can be answered, but cross-document questions cannot be accurately answered
Solution Approach 1:
The patent replaces conventional reading comprehension mechanisms with a knowledge graph-based approach. Instead of processing text sequentially through traditional NLP pipelines, the system extracts entities and relationships into structured knowledge graphs, then uses graph traversal and fusion to answer questions. This substitution enables precise cross-document answers while maintaining efficient processing through structured data representations.
Data Source
AI summary
Provided is a method, computer program product, and system for fusing knowledge graphs to generate a larger knowledgebase for responding to cross document questions. A processor may extract contextual information from a plurality of documents. The processor may generate, based on the extracted contextual information, a knowledge graph for each document of the plurality of documents. The processor may analyze each knowledge graph to determine if one or more entities of each knowledge graph are linked. The processor may fuse, in response to an entity in a first knowledge graph being linked to an entity in a second knowledge graph, the first knowledge graph with the second knowledge graph to create a fused knowledge graph.