Neural Net Content Recommendation for Personalized Screen Optimization

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing content recommendation systems for OTT video delivery and online music services often generate recommendation screens that are either irrelevant or redundant due to their reliance on human-curated content or similarity-based recommendations, failing to expose users to new relevant content.

Innovation Solution

A content recommendation application utilizing a neural net trained to incentivize the selection of diverse and explorative content while dis-incentivizing the selection of content already likely to be requested, by adjusting reward scores based on user interactions and refining neural connections to maximize relevance and diversity.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If content recommendations are curated by human editors, then content selection expertise is improved, but personalization for each individual user deteriorates

Engineering Contradiction:

Improvecontent selection expertiseVSAvoidpersonalization for each individual user

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent replaces the manual human editing process with an automated neural network system that processes user data and content features to generate personalized recommendations. The neural net learns from user interactions and automatically selects content, substituting the mechanical human curation process with an intelligent automated system that can adapt to individual user preferences at scale.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Adaptability or versatility

If groups of content items similar to most commonly requested content are displayed, then user preference alignment is improved, but content diversity deteriorates

Engineering Contradiction:

Improveuser preference alignmentVSAvoidcontent diversity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The system implements a feedback mechanism where the neural network receives signals from user interactions (clicks, views, skips) and adjusts its recommendations accordingly. The network learns from this feedback to balance showing content that aligns with user preferences while also introducing diverse content that the user might not have otherwise encountered, optimizing the trade-off between preference alignment and content diversity over time.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The recommendation system is dynamic and adaptive, continuously adjusting its output based on real-time user interactions and evolving preferences. Rather than providing static similar-content recommendations, the system dynamically modifies its suggestions to maintain an optimal mix of familiar content that aligns with user preferences and explorative content that introduces diversity, adapting as user behavior changes.

Inventive Principle:

Principle #15Dynamics

3Ease of manufacture

If traditional recommendation techniques are used, then implementation simplicity is improved, but effectiveness in providing new relevant content deteriorates

Engineering Contradiction:

Improveimplementation simplicityVSAvoideffectiveness in providing new relevant content

Core Design Contradiction:

Ease of manufactureVSProductivity

Solution Approach 1:

The patent replaces traditional recommendation techniques with a neural network-based system that leverages deep learning to understand user preferences and content characteristics. This substitution enables the system to provide more effective and personalized recommendations by learning complex patterns from data, overcoming the limitations of simpler traditional methods while delivering superior results in exposing users to new relevant content.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS20250047944A1Deep reinforcement learning for personalized screen content optimization

Publication Date: 2025.02.06 ADEIA GUIDES INC

AI summary

Systems and methods are described for selecting content item identifiers for display. The system may identify a set of content items that are likely to be requested in the future based on a history of content item requests. The system then selects a first plurality of content categories using a category selection neural net and selects a first set of recommended content items for the first plurality of content categories. The system increases a reward score for the first plurality of content categories based on receiving a request for a content item that is included in the first set of recommended content items. The system also decreases the reward score for the first plurality of content categories based on determining that the requested content item is included in the set of content items that are likely to be requested in the future. The neural net is trained based on the reward score of the first plurality of content categories to reinforce reward score maximization. The trained neural net is the used to select content items for display.