Streaming Platform Recommendation Ranking with Temporal BPR and LSTM
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional streaming platforms face challenges in recommending personalized content due to high user interaction complexity, time constraints, cold start problems, and lack of integration of temporal information and user interest alignment, leading to suboptimal viewer experience and retention.
Innovation Solution
A system and method that indexes user item interactions using Bayesian Personalized Ranking (BPR) and incorporates temporal information, employing LSTM optimization and cosine similarity to generate optimized recommendations based on user watch history and interaction data, utilizing techniques like Annoy for efficient indexing and sparse data handling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional systems recommend content based on user watch history and interactions, then personalization is provided, but time complexity increases and training time extends
Solution Approach 1:
The patent applies preliminary action by pre-computing and storing user-item interaction indices before Bayesian Personalized Ranking (BPR) processing. This pre-indexing of user watch history and interaction data allows the system to quickly retrieve and process relevant information during recommendation generation, significantly reducing the time complexity and training time while maintaining personalization quality
2Stability of the object's composition
If conventional systems strictly base recommendations on user profile, then consistency is maintained, but temporal information and user interest alignment are lost
Solution Approach 1:
The patent applies dynamics by integrating temporal information into the recommendation system through time-decay functions and sequential modeling. The system dynamically adjusts recommendation weights based on recency of interactions and evolves user profiles over time, allowing recommendations to adapt to changing user interests while maintaining consistency through structured data organization and processing frameworks
3Adaptability or versatility
If conventional systems explore new content based on user profile, then content discovery is enabled, but balance with user interest alignment is lacking
Solution Approach 1:
The patent applies parameter changes by using cosine similarity as a measurable parameter to quantify alignment between recommended content and user interests. The system calculates cosine similarity between user preference vectors and content vectors, providing a precise measurement of interest alignment while enabling content discovery through exploration of items with high but not maximum similarity scores, thus balancing discovery with alignment
4Quantity of substance
If conventional systems process vast user interaction data, then comprehensive analysis is achieved, but processing complexity increases
Solution Approach 1:
The patent applies taking out (extraction) by selectively extracting and indexing only the most relevant features and interactions from vast user data. The system extracts key interaction patterns, user preferences, and item attributes, storing them in optimized index structures that reduce processing complexity while preserving the essential information needed for accurate recommendations
Data Source
AI summary
The present disclosure provides a system and method for ranking recommendations in a streaming platform. The system uses temporal information in calculations of Bayesian Personalized Rankings (BPR) to consider a sequence of user interactions. The system incorporates indexing user-items prior to the BPR calculations for faster training. The indices generated through the BPR technique are used as inputs to a long short-term memory (LSTM) optimization model which help in better convergence to an optimization function. Further, the system uses a content-based filtering technique and a collaborative filtering technique for processing user items during various modes of user interactions. Hence, the system provides an improved and meaningful ranking of recommendations to users on streaming platforms.