Query Clustering for Context-Aware Item Recommendations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing recommendation systems for network sites and electronic catalogs fail to adequately identify and consider the contexts in which users tend to acquire combinations of items, leading to incomplete and context-insensitive item bundle recommendations.
Innovation Solution
The implementation of search session analysis and query clustering techniques to generate query clusters, item descriptor clusters, and item clusters, which are used to provide contextual and accurate item recommendations by analyzing user search patterns and correlating related items, brands, and demographics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If existing recommendation systems use basic user activity monitoring to generate item recommendations, then the system can provide simple recommendations, but the recommendations fail to capture contextual information about item combinations
Solution Approach 1:
The patent segments the recommendation system into multiple components: query cluster generation module, item descriptor cluster generation module, and item cluster generation module. Each module processes specific aspects of search data independently, allowing the system to capture contextual information about item combinations without creating a monolithic complex system. The segmentation enables modular processing of search queries, user behaviors, and item relationships.
Solution Approach 2:
The patent introduces a new dimension of analysis by generating query clusters from search query data. Instead of only analyzing direct user-item interactions, the system adds the dimension of search query patterns and co-occurrence relationships. This dimensional expansion allows capture of contextual information about item combinations that users search for together, even when they don't purchase both items.
2Loss of information
If the recommendation system uses basic item co-purchase data, then the system can generate bundle recommendations, but it fails to identify the contexts in which users acquire item combinations
Solution Approach 1:
The patent performs preliminary action by pre-generating query clusters and item descriptor clusters from historical search data before generating item recommendations. The system pre-processes search queries to identify co-occurrence patterns and contextual relationships between items. This preliminary clustering of queries and descriptors enables the system to understand contexts in which users search for item combinations, improving subsequent recommendation accuracy.
Solution Approach 2:
The patent introduces query clusters and item descriptor clusters as intermediary structures between raw search data and final item recommendations. These intermediaries capture contextual relationships by grouping related queries and descriptors. The query clusters serve as mediators that bridge user search behavior patterns with item relationships, enabling the system to infer contextual understanding of item combinations without directly observing all user purchase contexts.
3Reliability
If the system generates recommendations based on limited search history data, then the system can respond quickly, but the recommendations lack stability over time as items are added or removed
Solution Approach 1:
The patent performs preliminary action by pre-generating and storing query clusters and item descriptor clusters from historical search data. This pre-processing creates a stable foundation of contextual relationships that can be reused for recommendations. When items are added or removed from the catalog, the pre-generated clusters provide stable reference frameworks, reducing the need for complete re-analysis and improving recommendation stability over time.
Solution Approach 2:
The patent enables parameter changes by allowing the system to adapt cluster compositions as items are added or removed from the catalog. The query clusters and item descriptor clusters are designed to be flexible structures that can incorporate new items or remove obsolete items while maintaining overall cluster integrity. This parameter adaptability ensures recommendation stability without requiring complete re-generation of clusters, balancing reliability with efficient use of processing time.
Data Source
AI summary
The present disclosure provides computer-implemented systems and processes for clustering items and improving the utility of item recommendations. One process involves applying a clustering algorithm to users' search session queries over periods of time to generate query clusters comprising correlated query terms. Correlations may be based on, among other things, the frequency of which query term pairs appear together in a single search session. The generated query clusters may be used to generate item descriptor clusters indicative of items and/or types of items that may be complementary. Other criteria may be applied to the query and item clusters to generate variant clusters. For example, information such as related brands, market segments, and other data may be applied to item descriptor clusters to generate item clusters that include complementary items associated with or targeted for particular demographics. Item descriptor clusters and item clusters can be used to improve the item recommendations.