Digital Item Similarity Scoring via Interaction History
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional digital content distribution systems face inefficiencies in accurately recommending similar digital media due to the computational and storage-intensive nature of meta-data tags, which may not accurately describe content and vary in significance depending on the content being evaluated.
Innovation Solution
A method that determines global and local interaction counts across users to compute an interaction probability and similarity score, combining these with a popularity score to recommend digital content items based on viewing patterns, thereby enhancing accuracy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If meta-data tags are used to identify similar digital media, then digital media can be categorized and recommended, but the system becomes computationally and storage space intensive
Solution Approach 1:
The patent extracts the core similarity signal from interaction history data, separating it from the complex meta-data tag system. Instead of using thousands of tags, the invention computes similarity directly from user interaction patterns (views, purchases, ratings), extracting only the essential behavioral signals that indicate similarity between items.
Solution Approach 2:
The patent creates a simplified model of item similarity by copying and analyzing interaction patterns rather than copying and comparing extensive meta-data tags. The system builds a probabilistic model based on observed user behaviors, creating a lightweight representation that captures similarity without requiring heavy tag storage and processing.
2Measurement precision
If thousands of meta-data tags are used to accurately describe digital media contents, then similarity computation accuracy improves, but storage space and computational resources increase
Solution Approach 1:
The invention extracts the essential similarity information directly from user interaction data, eliminating the need to store and process thousands of meta-data tags. The system captures content similarity through observed user behaviors (viewing patterns, purchase sequences, rating correlations), extracting only the necessary signals for accurate recommendation.
Solution Approach 2:
The patent uses lightweight, dynamically computed interaction statistics instead of persistent, extensive meta-data tag structures. The similarity measures are computed on-demand from interaction history, using minimal storage for aggregated statistical data rather than maintaining large tag databases for each digital media item.
3Ease of manufacture
If meta-data tags are attached based on individual perception, then digital media can be categorized, but the tags may not accurately describe the contents and similarity computation becomes inaccurate
Solution Approach 1:
The patent uses real user interaction feedback (views, purchases, ratings, search queries) to objectively determine item similarity, replacing subjective meta-data tagging. The system continuously learns from actual user behavior patterns, allowing the similarity measures to be dynamically refined based on observed preferences and associations, ensuring accuracy reflects real user perceptions rather than curator assumptions.
4Measurement precision
If content-specific scoring functions are used to determine similarity, then accuracy improves, but the functions become difficult to estimate
Solution Approach 1:
The patent develops a universal probabilistic framework that handles diverse digital media types and user interactions through a single cohesive model. The same interaction history analysis and probability computation methods apply whether recommending movies, music, books, or other content, eliminating the need to create and estimate separate content-specific scoring functions while maintaining accuracy across different media types.
Data Source
AI summary
One embodiment sets forth technique for computing a similarity score between two digital items is computed based on interaction histories associated with global users and interaction histories associated with local users. Global counts indicating the number of interactions associated with each unique pair of digital items are weighted based on a mixing rate. The weighted global counts are then combined with local counts to compute total counts. An effective interaction probability indicating the likelihood of a user interacting with one digital item in the pair of digital items after interacting with the other digital item in the pair is computed based on the total counts. The effective interaction probability is then corrected for noise, resulting in a similarity score indicating the similarity between the pair of digital items.