Superphrase Indexing for Search Precision and Recall
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional automated search technologies struggle to identify relevant documents due to restrictive matching requirements, often missing relevant articles with title match searches while producing false positives with abstract or full text matches, leading to under-reporting and over-reporting of matches.
Innovation Solution
The technique involves identifying 'superphrases' of 'seed phrases' using a data structure similar to an inverted index, allowing for the rewriting of search queries to capture more documents by imposing different matching requirements on various subsets of terms, thereby enhancing search result relevance and coverage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If title match search is used, then search precision is improved, but search recall deteriorates (relevant documents are missed)
Solution Approach 1:
The patent segments the search query into multiple phrases and identifies superphrases by sorting and comparing phrase components. This allows the system to break down complex queries into manageable units while maintaining the ability to match variations, thereby improving recall without sacrificing precision.
Solution Approach 2:
The patent performs preliminary processing by sorting distinct seed phrase words and indexing them before actual search execution. This pre-processing creates an optimized data structure that enables faster and more comprehensive matching during search operations, capturing more relevant documents efficiently.
2Reliability
If abstract or full text match search is used, then search recall is improved, but search precision deteriorates (false positives increase)
Solution Approach 1:
The patent applies different matching strategies to different parts of the search query. By identifying superphrases and their component words, the system can apply stricter matching to critical terms while maintaining flexibility for other terms, thus reducing false positives while preserving recall.
Solution Approach 2:
The patent dynamically adjusts matching requirements based on the identified superphrases and their relationships. The system can adaptively apply different matching thresholds to different phrases within the query, optimizing the balance between recall and precision for each specific search context.
3Productivity
If conventional automated query parsing is used, then processing speed is improved, but concept understanding deteriorates (radically different queries representing same concept cannot be mapped)
Solution Approach 1:
The patent adds a new dimension to query processing by analyzing phrases at multiple levels (individual words, phrase components, superphrases). This multi-dimensional approach enables the system to recognize that radically different queries may share common superphrases, thereby improving concept understanding while maintaining efficient processing through the sorted index structure.
Data Source
AI summary
Methods and apparatus are described by which “superphrases” of “seed phrases” representing basic concepts may be identified without having to compare all possible pairs of seed and candidate phrases. According to one class of embodiments, a data structure similar to an inverted index is used for indexing phrases. The elimination of seed and candidate phrase pairs is enabled by building and traversing the index in a particular manner.