Generative AI Contextual Advertising with Scene-Based Creative Placement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing online advertising methods are inefficient and disruptive, lacking relevance to the content they accompany, leading to user dissatisfaction and reduced revenue for publishers.
Innovation Solution
A system that utilizes a multimodal metadata extraction and deep learning to understand video content on a scene-by-scene basis, enabling contextual advertising by identifying optimal moments for ad insertion and superimposition based on rich metadata, using AI techniques to enhance relevance and engagement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional online advertising methods are used, then advertisements can be delivered to users, but the ads are inefficient and disruptive, lacking relevance to the content
Solution Approach 1:
The video content is segmented into scenes, and advertisements are selectively placed at specific scene transitions rather than continuously. This segmentation allows ads to be relevant to specific content segments while minimizing disruption during other portions of the video.
Solution Approach 2:
Different portions of the video content are treated differently: some scenes have advertisements superimposed while others do not. The ad placement is localized to specific scenes where it is most relevant, creating local quality variations that improve relevance while reducing overall disruption.
2Productivity
If more advertisements are displayed to increase revenue, then advertiser value propositions improve, but user engagement decreases due to disruption
Solution Approach 1:
The advertisement placement is dynamic rather than static. The system analyzes video content in real-time and dynamically places ads at optimal moments based on scene transitions and content relevance, maximizing monetization while minimizing user engagement reduction.
Solution Approach 2:
The system uses feedback from content analysis to determine optimal ad placement. By continuously monitoring scene transitions and content characteristics, the system adjusts ad placement to maximize relevance and minimize disruption, improving the balance between revenue and user engagement.
3Productivity
If ads are placed at all times to maximize visibility, then advertising impact increases, but contextual relevance decreases
Solution Approach 1:
The system performs preliminary analysis of video content to identify scene transitions and optimal ad placement moments before ads are actually displayed. This preliminary action ensures that ads are only placed when contextual relevance is highest, maximizing both visibility and relevance.
Solution Approach 2:
The system changes the parameter of ad placement timing based on content characteristics. By adjusting when ads are displayed relative to scene transitions and content flow, the system optimizes the balance between visibility and contextual relevance.
4Adaptability or versatility
If manual ad placement is used to ensure relevance, then contextual accuracy improves, but system complexity and cost increase
Solution Approach 1:
The system performs self-service by automatically analyzing video content and determining optimal ad placement without manual intervention. The automated scene detection and relevance assessment reduce system complexity compared to manual processes while maintaining high contextual accuracy.
Solution Approach 2:
Manual mechanical processes of ad placement are replaced with automated computational systems that use algorithms to analyze content and determine placement. This substitution reduces operational complexity while improving consistency and accuracy of contextual relevance.
Data Source
AI summary
A system for contextual modification of content based on multimodal extraction of metadata from the content, wherein the metadata is extracted by processing one or more scenes in the content to extract metadata corresponding to multiple extraction modes, and an embedding model for each extraction mode wherein an aggregated embedding model responsive to the extracted metadata for each mode formulates an aggregated embedding. A process controller may include an embedding extractor responsive to a control input. The control input may specify one or more features appearing in the content defining a content modification opportunity. The embedding extractor may include an embedding model coordinated with the embedding model for one or more of the embedding modes to generate an opportunity embedding in the form of a vector. A vector comparison processor determines the distance between the opportunity embedding and the aggregated embedding, wherein the embeddings are in the form of vectors. The process controller is responsive to the vector comparison processor to generate edit control instructions indicating a modification of the content upon detection of the content modification opportunity. A content editor is responsive to the edit control instructions to modify the content.