Virtual Clickstream Recommendations for Sparse User Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing recommendation systems are less effective for users with little or no purchase and item selection history, as they rely on outdated data or popular items, failing to accurately reflect current user interests.
Innovation Solution
The system collects and analyzes conversation data from social networking sites and other collaborative content platforms to infer user interests, generating personalized recommendations by treating product references in conversations as virtual item selections and integrating this data with catalog activity data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If recommendation systems rely on purchase and item selection history, then recommendations can be generated for users with existing data, but new and infrequent users receive poor recommendations due to lack of data
Solution Approach 1:
The patent introduces conversation data from social networking sites as an intermediary data source. This mediator bridges the gap for users with insufficient catalog activity by providing alternative behavioral signals (conversations about products) that can be transformed into virtual clickstream data, enabling recommendation generation for new and infrequent users without requiring extensive purchase history
Solution Approach 2:
The system performs preliminary data collection from social networking sites before catalog activity data becomes available or sufficient. By collecting and processing conversation data in advance, the system creates virtual clickstream records that can immediately support recommendation generation for new users, rather than waiting for users to accumulate sufficient purchase history
2Reliability
If recommendation systems use outdated purchase history data, then they can maintain data availability, but the recommendations fail to reflect current user interests
Solution Approach 1:
The system dynamically switches between data sources based on user activity levels and data freshness. For active users, it uses current catalog activity data; for inactive or new users, it dynamically incorporates conversation data from social networking sites. This dynamic approach ensures recommendations always reflect current user interests regardless of catalog activity timing
Solution Approach 2:
The recommendation system becomes multi-functional by accepting multiple types of user behavioral data: traditional catalog activity data (purchases, views) and conversation data from social networking sites. This universality allows the system to serve both frequent catalog users and new/infrequent users effectively, maintaining recommendation relevance across different user segments and time periods
3Reliability
If the system collects conversation data from social networking sites, then recommendation accuracy improves for users with little catalog activity, but system complexity increases
Solution Approach 1:
The patent employs intermediary components including data collection modules that interface with social networking sites, data processing modules that analyze conversation data, and virtual clickstream generation modules that transform conversation data into recommendation-ready formats. These intermediaries manage the complexity of multi-source data integration while maintaining recommendation accuracy
Solution Approach 2:
The system creates virtual copies of clickstream data from conversation data. Instead of directly using raw conversation data in recommendation algorithms, it generates virtual clickstream records that mimic the structure and utility of actual catalog interaction data, allowing existing recommendation systems to work with augmented data without major structural changes
Data Source
AI summary
This disclosure describes various processes for collecting information about users from sources other than catalog activity data. This information can be used to generate recommendations for users with activity data deficiencies. Some example sources for this supplemental data include collaborative content sites, such as social networking sites. Social networking sites typically allow users to engage in conversations with other users through text, audio, and/or video. Conversation data collected from these sites or from other sources can be analyzed to infer user interests. A recommendation process (or other application) can use the inferred interests to supplement or take the place of catalog activity data.