Search System Term Weighting for Reduced Query Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional search systems often yield zero or few search results for lengthy or wordy queries, especially when additional unrelated words are included, leading to inadequate product or search results despite the existence of relevant information.
Innovation Solution
A search system and method that computes the weight of each term in a query using morpheme analysis and mutual information, removes less important terms based on computed weights, and generates a reduced query to improve search results by selecting more relevant terms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If a lengthy or wordy query is used, then the query may include more detailed information, but the number of search results decreases to zero or very few
Solution Approach 1:
The query is segmented into multiple terms using morpheme analysis, allowing the system to evaluate each term's contribution to the search. This segmentation enables identifying and removing less important terms while preserving the core search intent, thus maintaining information completeness while improving result quantity.
Solution Approach 2:
The system extracts and removes unnecessary or less important terms from the query based on computed weights. By taking out these redundant terms, the query becomes more effective, increasing the number of search results without losing essential search intent.
2Measurement precision
If additional words are included in the query, then the query may be more descriptive, but search results become insufficient or unrelated
Solution Approach 1:
The system applies different quality standards to different terms in the query. By computing individual weights for each term based on morpheme analysis and mutual information, the system identifies which terms contribute positively to relevance and which degrade it, allowing selective removal of harmful terms while preserving descriptive quality.
Solution Approach 2:
The system uses an intermediary mechanism (weight computation based on mutual information and query log analysis) to evaluate the impact of each term on search result relevance. This intermediary assessment allows the system to filter out terms that reduce reliability while maintaining those that enhance descriptiveness.
3Measurement precision
If a query includes category name, brand name, and manufacturer, then the query is very specific, but search results are zero or very few
Solution Approach 1:
The system applies partial action by selectively removing only the least important terms from the query rather than removing all specific terms. This allows maintaining adequate specificity while reducing excessive constraints that lead to zero results, achieving a balance between precision and productivity.
Data Source
AI summary
A search system includes a non-transitory computer-readable storage device, and a computing device. A term extracting unit causes the computing device to extract multiple terms from an input query, a weight computing unit causes the computing device to compute a weight for each of the multiple terms, and a reduction query recommending unit causes the computing device to remove at least one term from the multiple terms, based on the weight, and to provide at least one reduced query using remaining terms. A search method that uses a computing device to provide a reduced query, the method includes extracting multiple terms from an input query; computing, by the computing device, a weight of each of the multiple terms; and removing at least one term from the multiple terms, based on the weight, and providing at least one reduced query generated using remaining terms out of the multiple terms.