Document Search Using Metamodel Semantic Network
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing search techniques for similar documents, such as Latent Semantic Indexing (LSI), are inaccurate due to insufficient context information, particularly with technical documents that use abbreviations, leading to mismatches in context determination.
Innovation Solution
The use of a metamodel semantic network to analyze documents, which represents semantic relations among terms, concepts, and domains, allowing for the identification of common terms and their frequencies, and storing these in a searchable index to locate similar documents based on frequency comparisons.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If LSI is used to search for similar documents based on keyword matching, then the search process is simple and fast, but the accuracy is low due to insufficient context information and abbreviation mismatches
Solution Approach 1:
The patent introduces an external knowledge base as an intermediary between the search query and document collection. This knowledge base contains pre-established semantic relationships and abbreviation mappings that mediate the matching process, enabling accurate interpretation of abbreviations and contextual meanings without requiring complex analysis of the document texts themselves.
Solution Approach 2:
The patent performs preliminary action by pre-processing and storing semantic relationships, abbreviation expansions, and contextual information in an external knowledge base before the actual search operation. This advance preparation allows the search system to quickly retrieve accurate semantic mappings during query processing, improving both accuracy and efficiency.
2Measurement precision
If more context information is extracted from documents to improve search accuracy, then the determination of context becomes more accurate, but the processing time and computational resources increase
Solution Approach 1:
The patent extracts and stores context information, semantic relationships, and abbreviation mappings in advance in an external knowledge base. During search operations, the system simply queries this pre-processed knowledge base rather than performing complex context analysis on each document, significantly reducing processing time while maintaining high accuracy.
Solution Approach 2:
The patent extracts essential context information and semantic relationships from documents and stores them in an external knowledge base, separating this preprocessing step from the actual search operation. This extraction allows the search system to work with compact, pre-analyzed data structures rather than raw documents, improving search speed.
Data Source
AI summary
In an embodiment, a method is provided for searching similar documents. Here, a document is accessed and terms from a metamodel semantic network is identified. The document is analyzed to identify a number of the terms from the metamodel semantic network that are also found in the document, and to identify a frequency of occurrence in the document for each term. A search is conducted for other documents having frequencies of occurrences that are similar to the identified frequency of occurrence. These other documents have been previously analyzed using the same terms from the metamodel semantic network.