Video Game Content Identification via Machine Learning Embeddings

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional methods for identifying interesting content or engaging players in esports and video games rely heavily on human annotation and profiling, which are time-consuming and not scalable, and lack robustness in content analysis.

Innovation Solution

A system using machine learning for content analysis, specifically embedding representation learning and time sequence analysis, to automatically identify and recommend portions of video game content based on popularity and similarity, reducing the need for human input and annotation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of manufacture

If human annotation and profiling methods are used to identify interesting content and group players, then the content analysis can be performed with simple tools, but the process becomes time-consuming and not scalable

Engineering Contradiction:

Improveease of content analysisVSAvoidscalability of content analysis

Core Design Contradiction:

Ease of manufactureVSProductivity

Solution Approach 1:

The patent replaces manual human annotation and profiling with automated machine learning systems. Content analysis is performed using embedding representation learning and time sequence analysis algorithms that automatically process video game footage, eliminating the need for human reviewers while dramatically increasing processing speed and scalability.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system enables content to analyze and tag itself automatically through machine learning models. The embedding representation learning extracts features from video content, and time sequence analysis automatically identifies interesting segments based on learned patterns, allowing the system to self-service without human intervention.

Inventive Principle:

Principle #25Self-service

2Ease of manufacture

If human annotation is used for content identification, then the analysis can be performed with simple tools, but robustness in content analysis is lacking

Engineering Contradiction:

Improvesimplicity of analysis methodVSAvoidrobustness of content analysis

Core Design Contradiction:

Ease of manufactureVSReliability

Solution Approach 1:

The patent replaces unreliable human annotation with robust machine learning-based content analysis. The embedding representation learning and time sequence analysis provide consistent, reproducible results that are not subject to human error, fatigue, or subjective bias, thereby significantly improving the reliability and robustness of content identification.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Reliability

If extensive human annotation is performed to identify engaging content, then content quality can be ensured, but the process requires extensive human input and is not scalable

Engineering Contradiction:

Improvequality of content identificationVSAvoidamount of human input required

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent substitutes extensive human annotation efforts with automated machine learning systems that perform content analysis at scale. The embedding representation learning extracts meaningful features from video content, and time sequence analysis identifies engaging segments automatically, maintaining quality while eliminating the need for extensive human input.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system changes the parameters of content analysis by using learned embeddings and temporal patterns instead of manual evaluation criteria. This allows the system to automatically adjust to different content types and user preferences while maintaining high quality identification without human intervention.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS11654372B2Methods, systems, and devices for identifying a portion of video content from a video game for a player or spectator

Publication Date: 2023.05.23 AT&T INTELLECTUAL PROPERTY I L P

AI summary

Aspects of the subject disclosure may include, for example, obtaining portions of video content from a video game from video game server(s) associated with a video game provider, selecting a first portion of video content from the portions of the video content, and providing the first portion to device(s) associated with viewer(s). Each device presents the first portion of the video content. Further embodiments include obtaining popularity information from the device(s) according to feedback based on presenting the first portion of the video content to the device(s), determining that the popularity information satisfies a popularity threshold associated with the video content, determining a subject matter corresponding to the first portion of the video content, and identifying a second portion of the video content from the video game to be recorded according to the subject matter. Other embodiments are disclosed.