Synonym Discovery Framework Using Query Logs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Information retrieval systems often fail to identify desired information due to discrepancies between user queries and the canonical names used by the system, as standard thesauruses are inadequate for handling query expansion tasks, particularly in generating synonyms.
Innovation Solution
A similarity analysis framework that leverages multiple similarity analysis functions, including pseudo-document and query context analysis, to generate synonyms that satisfy core properties such as symmetry, strength-of-similarity, and relatedness-of-class, using query log data and indexing mechanisms to expedite computations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If standard thesaurus resources are used for query expansion, then the system can generate synonyms, but the synonyms fail to capture user intent accurately for complex queries
Solution Approach 1:
The patent changes the parameters for synonym generation by using multiple similarity analysis functions with different weighting parameters. Each function (e.g., query context similarity, pseudo-document similarity) uses specific parameters to evaluate different aspects of synonymhood, allowing the system to adapt to various query types and generate accurate synonyms that standard thesauruses cannot produce.
Solution Approach 2:
The patent creates a composite evaluation framework by combining multiple similarity analysis functions into a unified synonym generation system. This composite approach integrates query context analysis, pseudo-document similarity, and other analysis functions to produce comprehensive synonym results that are more accurate and versatile than any single function could achieve alone.
2Reliability
If multiple similarity analysis functions are used to generate synonyms, then synonym accuracy improves, but computational resource consumption increases
Solution Approach 1:
The patent segments the synonym generation task into multiple independent similarity analysis functions, each responsible for specific aspects of evaluation. This segmentation allows the system to process queries in parallel using different functions simultaneously, reducing overall computational time while maintaining high accuracy through the combined results of all functions.
Solution Approach 2:
The patent applies partial action by using multiple similarity analysis functions that may not individually evaluate all possible synonym properties, but collectively they provide comprehensive coverage. The system performs sufficient analysis to achieve accurate synonyms without requiring exhaustive evaluation of every possible aspect, balancing computational resources with required accuracy.
3Reliability
If query log data is used for synonym generation, then synonyms reflect actual user behavior, but the system becomes more complex
Solution Approach 1:
The patent introduces pseudo-documents as intermediary structures that bridge query log data and synonym generation. These pseudo-documents represent user behavior patterns and query contexts in a structured format that can be efficiently processed by similarity analysis functions, simplifying the integration of query log data into the synonym generation process without requiring direct complex analysis of raw log data.
Data Source
AI summary
A similarity analysis framework is described herein which leverages two or more similarity analysis functions to generate synonyms for an entity reference string re. The functions are selected such that the synonyms that are generated by the framework satisfy a core set of synonym-related properties. The functions operate by leveraging query log data. One similarity analysis function takes into consideration the strength of similarity between a particular candidate string se and an entity reference string re even in the presence of sparse query log data, while another function takes into account the classes of se and re. The framework also provides indexing mechanisms that expedite its computations. The framework also provides a reduction module for converting long entity reference strings into shorter strings, where each shorter string (if found) contains a subset of the terms in its longer counterpart.