Latent Interest Modeling for ID-Less Media Content Targeting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems fail to effectively target media content to users without relying on traditional identifiers like cookies or deterministic IDs due to data privacy concerns and regulations, leading to poor targeting outcomes and limited personalization capabilities.
Innovation Solution
Latent interest models leverage context from the bidstream to build a propensity network, identifying co-occurring content signals and personalizing media content without relying on explicit user identifiers, allowing for ID-less targeting and expanding reach through latent propensities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional identifiers (cookies, MAIDs, IP addresses) are used for user identification and targeting, then targeting precision and personalization are improved, but data privacy concerns and regulatory compliance deteriorate
Solution Approach 1:
The patent extracts and removes traditional deterministic identifiers (cookies, MAIDs, IP addresses) from the targeting system. Instead of relying on these explicit user identifiers, the system uses contextual signals and latent interest models to achieve targeting without personally identifiable information, thereby resolving the contradiction between targeting precision and data privacy concerns
Solution Approach 2:
The patent introduces contextual signals and latent interest models as intermediaries between user behavior and advertising delivery. These intermediaries enable targeting by inferring user interests from contextual data rather than directly identifying users, maintaining targeting effectiveness while addressing privacy concerns
2Object-affected harmful factors
If traditional deterministic identifiers are eliminated for privacy compliance, then data privacy protection is improved, but targeting effectiveness and audience addressability deteriorate
Solution Approach 1:
The patent replaces the mechanical identifier-matching system with a machine learning-based latent interest model. Instead of mechanically matching users to ads through deterministic IDs, the system uses AI models to predict user interests from contextual signals, maintaining targeting effectiveness without requiring traditional identifiers
Solution Approach 2:
The patent changes the fundamental parameters of targeting from deterministic identifier-based matching to probabilistic interest-based matching. By transforming the targeting approach from exact matches on user IDs to statistical predictions of user interests, the system maintains effectiveness while enabling privacy compliance
3Object-affected harmful factors
If cross-site tracking is restricted for privacy reasons, then user privacy is protected, but audience segmentation and personalization capabilities deteriorate
Solution Approach 1:
The patent segments the targeting approach into two distinct components: contextual signal collection (which does not require cross-site tracking) and latent interest model processing (which creates audience segments based on inferred interests rather than tracked behavior). This segmentation enables privacy-protected audience segmentation by replacing cross-site tracking with contextual analysis
Solution Approach 2:
The patent creates copies of user interest profiles through latent interest models without copying actual user identifiers or tracking data. The system generates synthetic representations of user interests based on contextual signals, enabling personalization capabilities while protecting user privacy by never copying or storing actual personal information
Data Source
AI summary
A method and system deliver media content. Multiple first requests for media content are received and first key phrases are extracted into a dictionary. Second media content requests and associated context are collected for media content served to a latent interest audience. The context is used to obtain candidate key phrases. For each of the contexts, the candidate key phrases that match the first key phrases are determined. A subset of the matching candidate key phrases are assigned to the campaign. The subset of matching candidate key phrases is expanded using semantic similarity and stored in a second table. A new media content request that does not include an explicit user identification is received. Based on the new media content request, new request key phrases are determined and compared to the expanded subset. Campaigns with intersections are selected and utilized to deliver media content.