Relevancy Score Calculation for E-Commerce Product Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current search engines struggle to effectively rank the relevancy of search results for electronic commerce websites, as existing relevancy scoring models are not optimized for product searches, leading to irrelevant results being obscured by relevant ones due to reliance on frequency of query terms in general text document searches.
Innovation Solution
A method for calculating relevancy scores based on match properties such as the number of matched terms, match length, query length, and linguistic factors, using specific formulas for pattern and conventional tokenized text searches, including boost and penalty factors to prioritize relevant results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If frequency-based relevancy scoring models are used for general text document searches, then the search engine can process queries efficiently, but the relevancy scoring is not optimized for product searches causing relevant results to be obscured by irrelevant ones
Solution Approach 1:
The patent applies local quality by differentiating the scoring approach for different fields (e.g., product attributes vs. general description). Each field is assigned specific weights and scoring parameters tailored to its nature, allowing the system to optimize relevancy scoring for product-specific fields while maintaining efficiency for general fields.
Solution Approach 2:
The system changes the parameter set used for relevancy scoring from generic frequency-based parameters to field-specific parameters including match properties, field weights, and query-term importance factors. This parameter transformation enables the same search engine to handle both general text searches and product-specific searches effectively.
2Measurement precision
If existing relevancy scoring models are used, then the search engine maintains simplicity in processing, but the precision and recall of search results deteriorate for electronic commerce websites
Solution Approach 1:
The relevancy scoring process is segmented into distinct calculation stages: query processing, field matching, score calculation, and result ranking. Each stage handles specific computations independently, making the overall complex scoring process manageable and executable through systematic breakdown of the calculation tasks.
Solution Approach 2:
The system performs preliminary actions by pre-calculating and storing field weights, query-term importance factors, and match property thresholds before actual search execution. This preparation reduces the computational burden during live search queries while maintaining high precision in relevancy scoring.
3Reliability
If generic text search formulas are applied to product searches, then the search system remains simple to implement, but the recall of relevant product results is reduced
Solution Approach 1:
The patent creates a universal relevancy scoring framework that can handle both general text searches and product-specific searches using the same core system. The framework incorporates multiple scoring components (field weights, match properties, query-term factors) that can be activated or adjusted based on the search type, eliminating the need for separate search engines while maintaining high recall for product searches.
Data Source
AI summary
A method including the steps of: receiving, using one or more processors, a search query containing search terms from a user device; executing, using one or more processors, a query on a database based in response to the received search query; generating, using or more processors, search results based on the executed query; calculating, using one or more processors, a relevancy score for each of the generated search results, wherein the relevancy score is determined based on match properties of the search terms and field terms of a matching field; assigning, using one or more processors, a calculated relevancy score to each of the generated search results; and transmitting, using or more processors, the scored search results for display as a ranked list on the user device.