Multi-hop Entity Relationship Search via Graph Centrality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The computational complexity of analyzing groups of documents to discover relationships between two entities increases rapidly with the number of intermediate entities required to substantiate the relationship, making it challenging to efficiently identify and substantiate multi-hop relationships in entity-based systems.
Innovation Solution
An unsupervised multi-hop search strategy is employed, which uses graph centrality to pre-select important entities, combined with A* search and beam search algorithms to efficiently traverse the entity relationship graph and identify multi-hop paths between entities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional document analysis methods are used to discover relationships between entities, then relationship discovery is possible, but computational complexity increases rapidly with the number of intermediate entities
Solution Approach 1:
The patent segments the document analysis task into multiple stages: first identifying candidate intermediate entities through graph centrality analysis, then performing targeted document analysis only for promising candidates. This segmentation reduces the overall computational complexity by avoiding exhaustive analysis of all possible intermediate entities while maintaining relationship discovery accuracy.
Solution Approach 2:
The patent applies preliminary action by using graph centrality measures to pre-rank and select the most important intermediate entities before performing detailed document analysis. This preliminary filtering step identifies the most likely candidates for relationship substantiation, reducing the search space and computational resources required for subsequent analysis.
2Reliability
If exhaustive search through all intermediate entities is performed, then complete relationship substantiation is achieved, but memory requirements and processing time increase significantly
Solution Approach 1:
The patent applies local quality by focusing computational resources on the most promising intermediate entities identified through graph centrality analysis, rather than uniformly processing all entities. This localized approach concentrates memory and processing resources on the most relevant parts of the entity relationship graph, reducing overall memory requirements while maintaining substantiation completeness.
Solution Approach 2:
The patent uses graph centrality analysis as an intermediary mechanism that bridges the gap between exhaustive search requirements and memory constraints. By using centrality measures as a filtering intermediary, the system identifies a manageable subset of candidate entities that are most likely to substantiate relationships, reducing memory requirements while maintaining search thoroughness.
3Adaptability or versatility
If multi-hop search is performed to find relationships through intermediate entities, then relationship discovery capability is improved, but the number of search hops increases computational burden
Solution Approach 1:
The patent applies dynamics by making the search depth adaptive rather than fixed. The system dynamically determines how many hops to search based on the quality of intermediate entities identified through graph centrality analysis and the strength of evidence found in documents. This dynamic adjustment allows the system to achieve high relationship discovery capability while maintaining search efficiency by stopping the search when sufficient evidence is found or when additional hops yield diminishing returns.
Data Source
AI summary
An unsupervised multi-hop search across a corpus of documents in a database or other data resource permits the identification of relationships between two entities mentioned in the corpus in cases where the two entities are not co-mentioned within any documents in the corpus (or not mentioned within document(s) with sufficient frequency or proximity to infer the relationship). The search can employ a beam search algorithm anchored by word embeddings and an A* graph traversal to calculate semantic distance between the entities as different paths through the corpus for different entity co-mentions are evaluated.