Query-ASIN Association Graph for Search Precision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current search paradigms face challenges in retrieving relevant information due to the proliferation of online content, leading to inefficient recall and ranking of search results, especially when queries use different terminology than product descriptions.
Innovation Solution
The use of behavioral signals, such as user interactions and metadata, to create a query-ASIN association graph that weights and ranks search results more accurately, allowing for improved recall and reduced defective results by analyzing individual signal contributions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional search paradigms are used to retrieve information from online content, then the search system is simple to implement, but the search result recall and ranking precision deteriorates due to terminology mismatches between queries and product descriptions
Solution Approach 1:
The patent segments user behavior data into multiple distinct behavioral signals (clicks, purchases, add-to-cart, wish-list, views) and processes each signal type separately through individual association graphs. This segmentation allows the system to capture different aspects of user intent with specialized graphs, improving search precision while managing complexity through modular processing of each signal type independently.
Solution Approach 2:
The patent transitions from traditional single-dimension keyword matching to a multi-dimensional approach by incorporating behavioral signals as additional dimensions for associating queries with products. The system creates association graphs that operate in this extended dimensional space, where products are connected not only by textual keywords but also by multiple behavioral dimensions, thereby improving recall precision without proportionally increasing system complexity.
2Reliability
If multiple behavioral signals are aggregated to improve search results, then search result relevance improves, but the difficulty of detecting and measuring individual signal contributions increases
Solution Approach 1:
The patent creates separate association graphs for each behavioral signal type (clicks, purchases, add-to-cart, wish-list, views), allowing the system to measure and detect the contribution of each signal independently. By segmenting the analysis into distinct signal-specific graphs, the system can evaluate the individual impact of each behavioral signal on search results while maintaining overall relevance through aggregation of all signal contributions.
Solution Approach 2:
The system uses behavioral signals as feedback mechanisms where user actions (clicks, purchases, etc.) provide information that continuously refines the association graphs. Each behavioral signal serves as feedback that adjusts the relevance scoring, allowing the system to detect and measure signal contributions through the strength and pattern of these feedback loops in the association data.
3Measurement precision
If the search system analyzes individual behavioral signals separately, then the precision of matching user intent improves, but the processing time and computational resources increase
Solution Approach 1:
The patent pre-computes association graphs for each behavioral signal type and stores them for rapid retrieval during search operations. By performing the computationally intensive analysis of individual behavioral signals in advance and caching the results in association graphs, the system achieves high user intent matching precision while minimizing real-time processing time, as the actual search only requires querying pre-built association structures rather than analyzing raw behavioral data from scratch.
Data Source
AI summary
Systems and methods are disclosed for optimizing responses to queries. Analyses of user interactions and other behaviors can lead to association of queries with signals, including ASINs and other product descriptions. The associations can be algorithmically graphed and analyzed on a disaggregated basis and individually weighted to improve search recall and reduce the risk of returning defective search results. Machine learning techniques can further optimize the associations and/or the search results.