Video Viewing Spike Attribution via Textual Content Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Video providers and marketers face challenges in determining the reasons behind viewing spikes within videos, as existing analytics tools do not effectively identify the textual content responsible for these increases in user engagement.
Innovation Solution
A digital medium environment is developed to analyze video analytics data, detect viewing spikes, and identify textual content from video sources or referral sources that users have read prior to watching the video, using text analysis to determine the content's contribution to the spike, and calculate a contribution score for each instance of textual content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If video analytics data is analyzed to identify viewing spikes, then understanding of user engagement patterns is improved, but the ability to identify specific textual content responsible for spikes remains insufficient
Solution Approach 1:
The system segments the video content into discrete textual elements (transcripts, titles, descriptions, tags) and analyzes each segment's contribution to viewing spikes independently. This allows identification of specific textual content responsible for engagement without analyzing the entire video as a monolithic unit.
Solution Approach 2:
The patent introduces an intermediary text analysis system that bridges the gap between video analytics data and textual content. This intermediary layer processes video analytics data, correlates it with textual elements, and identifies causal relationships without requiring direct modification of the video playback system.
2Measurement precision
If text analysis is performed on video sources and referral sources, then identification of content responsible for viewing spikes is improved, but processing time and computational resources increase
Solution Approach 1:
The system performs preliminary actions by pre-processing and indexing textual content from video sources and referral sources before analytics are needed. Textual elements are segmented, tagged, and stored in an accessible format that enables rapid correlation with viewing analytics data when spikes occur.
Solution Approach 2:
Instead of analyzing all textual content uniformly, the system applies local quality by focusing analysis resources on specific textual elements that are most likely to cause viewing spikes based on their proximity to spike events and their engagement potential. Different weights and analysis depths are applied to different text segments.
3Loss of information
If contribution scores are calculated for multiple instances of textual content, then understanding of content impact is improved, but system complexity and computational load increase
Solution Approach 1:
The system changes parameters by using multiple scoring dimensions (viewing spike correlation, temporal proximity, engagement metrics) to evaluate textual content. Each textual element is assessed against multiple parameters simultaneously, and the results are aggregated into a single contribution score that reflects overall impact.
Solution Approach 2:
The patent uses copying by creating simplified representations or proxies for complex analytical relationships. Instead of modeling all possible interactions between textual content and viewer behavior, the system creates a copied simplified model that captures the essential contribution relationships through contribution scores.
Data Source
AI summary
A digital medium environment is described to determine textual content that is responsible for causing a viewing spike within a video. Video analytics data associated with a video is queried. The video analytics data identifies a number of previous user viewings at various locations within the video. A viewing spike within the video is detected using the video analytics data. The viewing spike corresponds to an increase in the number of previous user viewings of the video that begins at a particular location within the video. Then, text of one or more video sources or video referral sources read by users prior to viewing the video from the particular location within the video is analyzed to identify textual content that is at least partially responsible for causing the viewing spike.