Grammatical Unit Search for Document Retrieval Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing search techniques within digital documents are inefficient and prone to false positives and false negatives due to string-based matching, failing to effectively retrieve relevant information, especially when synonyms or contextual meanings are involved.
Innovation Solution
The system employs grammatical unit-based searching, where queries are processed to identify and rank relevant phrases, clauses, or sentences within documents, using feature extraction and machine learning techniques to provide accurate and relevant search results, even when exact terms are not present.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If string-based matching is used for searching, then the search process is simple and fast, but the search accuracy deteriorates due to false positives and false negatives
Solution Approach 1:
The patent segments the search process into multiple stages: initial string-based filtering followed by grammatical unit analysis. Documents are first filtered using quick string matching, then candidate results are re-evaluated using grammatical unit extraction and scoring, separating the speed advantage of string matching from the accuracy advantage of semantic analysis.
Solution Approach 2:
The patent introduces grammatical units (phrases, clauses, sentences) as an intermediary layer between raw string matching and final search results. These grammatical units serve as mediators that capture contextual meaning while building upon the initial string match results, allowing the system to maintain both speed and accuracy.
2Adaptability or versatility
If common words are searched using string matching, then the search covers broad topics, but the number of irrelevant results increases causing false positives
Solution Approach 1:
The patent applies local quality by analyzing the contextual quality of each search result individually. Instead of treating all matches equally, the system extracts grammatical units from each candidate document and calculates relevance scores based on local contextual features, allowing common words to maintain broad coverage while individual results are evaluated for genuine relevance.
Solution Approach 2:
The system implements feedback by using grammatical unit analysis to re-evaluate initial search results. The relevance scoring mechanism provides feedback on the quality of each match, allowing the system to filter out false positives while preserving true positives, even when searching for common words.
3Productivity
If exact character strings are required for matching, then the search is precise and fast, but relevant documents are missed due to synonym usage
Solution Approach 1:
The patent performs preliminary string-based filtering to quickly identify candidate documents, then applies grammatical unit analysis as a secondary pass. This preliminary action approach allows the system to maintain search efficiency while recovering relevant information that would be missed by strict string matching alone.
Solution Approach 2:
The system adds another dimension to the search by moving from one-dimensional exact string matching to multi-dimensional analysis that includes grammatical structure, phrase context, and semantic relevance. This dimensional expansion allows the system to find relevant documents using synonyms while maintaining the efficiency of exact matching for initial filtering.
Data Source
AI summary
Systems and techniques for searching within a document include receiving a query by way of a user interface of an application, and in conjunction with identification of the at least one document. A feature value characterizing a relevance of each grammatical unit of the document to the query may be extracted. The grammatical units may be ranked, based on each feature value of each grammatical unit. At least one selected grammatical unit of the plurality of grammatical units may then be displayed, based on the ranking.