Concept Graph Relevance Computation for Semantic Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional information retrieval technologies, such as keyword-based indexing and latent semantic analysis, fail to effectively utilize large volumes of crowd-sourced data and struggle with providing fast response times, especially when dealing with rich ontologies and unstructured data sources.
Innovation Solution
A method for conceptual analysis that uses a concept graph to compute the relevance of documents to concepts not explicitly mentioned, by accessing a concept graph with nodes representing concepts and edges representing relations, and combining a priori information from documents with the graph to generate posteriori information on concept relationships.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If query expansion techniques are applied to improve search quality by substituting search terms with synonyms and performing independent searches, then the quality of returned results is improved, but the computational cost and response time deteriorate significantly
Solution Approach 1:
The patent pre-computes and stores synonym relationships, concept hierarchies, and search result mappings in a knowledge base before actual search queries are executed. When a search is performed, the system retrieves pre-computed results rather than performing expensive independent searches for each synonym, thus maintaining high search quality while dramatically reducing response time.
Solution Approach 2:
The patent combines multiple search results from different query expansions into a unified ranked list using a scoring mechanism that aggregates results from synonym searches, concept hierarchy searches, and original query searches. This merging process eliminates redundant computations and provides comprehensive results efficiently.
2Adaptability or versatility
If query expansion techniques are applied with rich ontologies to improve search comprehensiveness, then more relevant results are found, but the number of searches increases from one to potentially tens of thousands
Solution Approach 1:
The patent extracts and stores the expansion query structure, synonym relationships, and concept hierarchies in a pre-built knowledge base. During actual search, only the necessary expansions are retrieved and executed rather than generating and executing all possible expansions, reducing the search complexity from potentially tens of thousands to a manageable number while maintaining comprehensiveness.
Solution Approach 2:
The patent implements a threshold-based filtering mechanism that limits the number of synonym expansions and concept hierarchies processed for each query. By setting reasonable thresholds on the number of synonyms to expand and concepts to traverse, the system achieves sufficient search comprehensiveness without the exponential explosion of search operations.
3Speed
If traditional keyword-based indexing is used to ensure fast response times, then search speed is maintained, but documents containing concepts not explicitly mentioned in keywords are not retrieved
Solution Approach 1:
The patent introduces a concept hierarchy and synonym network as an intermediary layer between keywords and documents. When a keyword search is performed, the system automatically expands the query through the concept hierarchy and synonym relationships to find documents that mention related concepts, thereby improving concept retrieval accuracy while maintaining the speed advantage of keyword-based indexing.
4Measurement precision
If latent semantic analysis is used to project documents to latent semantic space to improve concept matching, then abstract concepts can be represented, but the technique cannot effectively utilize crowd-sourced data from websites like Wikipedia
Solution Approach 1:
The patent creates a unified concept space that integrates traditional latent semantic analysis with crowd-sourced data from Wikipedia and other websites. The system processes both structured and unstructured data sources through a common conceptual framework, allowing abstract concept representation while effectively utilizing diverse data sources including crowd-sourced content.
Data Source
AI summary
According to an aspect, conceptual analysis of a document includes accessing a concept graph that includes a plurality of nodes and edges. Each node represents a concept and each edge represents a known relation between two concepts. Conceptual analysis of the document further includes computing a relevance of the document to concepts in the concept graph. The computing includes receiving a priori information about the document including concepts extracted from the document. The concepts extracted from the document include a subset of the concepts in the concept graph. The computing also includes combining the a priori information and the concept graph to generate a posteriori information that indicates a likelihood that the document is related to each of the concepts in the concept graph.