Hybrid Video Recommendation Engine Combining Collaborative and Content Filtering

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video recommendation systems fail to accurately suggest content to users with diverse interests, as they either rely on general trends, video-to-video similarities, or user consumption history, which can be limited in capturing current interests or providing variety.

Innovation Solution

A machine learning engine combines collaborative filtering and content-based filtering to predict user interest in videos by weighing various input parameters, including user behavior and video features, to generate personalized recommendations and provide explanations for the suggested content.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If collaborative filtering is used to recommend videos based on user consumption history, then recommendations are personalized to user interests, but accuracy deteriorates when user history is insufficient or does not reflect current interests

Engineering Contradiction:
Improvepersonalization to user interestsVSAvoidrecommendation accuracy
Core Design Contradiction:
Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent combines collaborative filtering (user-to-video approach) with content-based filtering (video-to-video approach) into a hybrid recommendation system. This merging allows the system to leverage both user consumption history and video content features, ensuring accurate recommendations even when user history is limited or outdated. The system weighs multiple signal types including user behavior, video characteristics, and content similarity to produce robust recommendations.

Inventive Principle:
Principle #5Merging (Combining)

2Measurement precision

If content-based filtering is used to recommend similar videos, then recommendations are provided when user history is limited, but accuracy deteriorates when user interests are diverse

Engineering Contradiction:
Improverecommendation accuracyVSAvoidhandling diverse user interests
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent merges content-based filtering with collaborative filtering to create a hybrid system that overcomes the limitations of content-based filtering alone. While content-based filtering provides accurate recommendations for users with narrow interests or limited history, the integration with collaborative filtering and multiple signal types enables the system to adapt to users with diverse interests by incorporating population trends and broader consumption patterns.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The system dynamically adjusts the weighting of different recommendation signals based on user context. When a user has extensive and consistent viewing history, the system places higher weight on collaborative filtering signals. When user history is limited or shows diverse interests, the system increases weight on content-based filtering and trending signals, allowing flexible adaptation to different user scenarios.

Inventive Principle:
Principle #15Dynamics

3Measurement precision

If multiple recommendation signals are combined to improve accuracy, then recommendation quality increases, but system complexity increases

Engineering Contradiction:
Improverecommendation accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the recommendation system into distinct modular components: a first module that generates recommendations based on user consumption history (collaborative filtering), and a second module that generates recommendations based on video content features (content-based filtering). Each module processes specific signal types independently, and their results are combined through a weighting mechanism. This segmentation reduces overall system complexity by creating manageable, independent processing units with clear interfaces.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP2817970B1Automatically recommending content
Publication Date: 2022.12.21 OOYALA INC
  • EP2817970B1 patent drawingFigure 1
  • EP2817970B1 patent drawingFigure 2
  • EP2817970B1 patent drawingFigure 3

AI summary

Techniques are provided for selecting which videos to recommend to users by predicting the degree to which recommending each video will satisfy certain goals. To make the predictions, a trained machine learning engine is fed both collaborative filtering parameter values and content-based filtering parameter values. In the case of video-to-video recommendations, the collaborative filtering parameter values may be based on a video pair that includes a video in which a user has already demonstrated an interest. The machine learning engine generates a machine-learning score for each video. The machine learning scores are used as the basis for selecting which videos to recommend to a particular user.