Audience Selection via Embedding Space Semantic Ranking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audience selection methods for targeted content delivery, such as advertising, rely on coarse or misleading descriptive names, lacking precision and failing to account for users' diverse interests, leading to ineffective content targeting.
Innovation Solution
A method utilizing an embedding space to position search queries and audience records, along with associated keywords, allowing for precise ranking based on semantic similarity, enabling more accurate audience selection by analyzing behavioral data and browsing history.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If descriptive names are used to identify audiences, then the audience selection process is simple, but the precision and accuracy of audience identification deteriorates
Solution Approach 1:
The patent introduces an embedding space as an intermediary representation that bridges the gap between simple descriptive names and precise audience identification. The embedding space transforms audience records into vector representations that capture semantic meaning, allowing for both ease of operation through intuitive searching and high precision through semantic similarity calculations.
Solution Approach 2:
The patent changes the representation parameters of audience records from simple descriptive strings to multi-dimensional vector embeddings. This parameter transformation enables the system to capture nuanced semantic relationships and user behavior patterns, thereby improving identification precision while maintaining operational simplicity through standardized embedding operations.
2Device complexity
If coarse-grained descriptive titles are used for audiences, then the system complexity is low, but the effectiveness of content targeting deteriorates
Solution Approach 1:
The patent performs preliminary action by pre-computing and storing embedding representations for audience records before content targeting is needed. This allows the system to maintain low operational complexity during actual targeting tasks while achieving high effectiveness through pre-processed semantic representations that enable rapid and accurate audience matching.
Solution Approach 2:
The patent replaces mechanical string-matching systems with semantic embedding-based systems. Instead of relying on simple text comparison of descriptive titles, the system uses vector similarity calculations that capture semantic meaning, thereby improving content targeting effectiveness without proportionally increasing system complexity.
3Reliability
If proprietary audience composition data is kept private, then data security is maintained, but the advertiser's ability to accurately select audiences deteriorates
Solution Approach 1:
The patent extracts and utilizes only the necessary semantic information from proprietary audience data to create embedding representations. By taking out only the essential features needed for accurate targeting (such as interest patterns and behavioral characteristics) while discarding unnecessary detailed information, the system maintains data security through minimal data exposure while achieving high audience selection accuracy.
Data Source
AI summary
In some embodiments, a method includes determining a position for a search query and a position for each audience record from multiple audience records in an embedding space. The method further includes receiving multiple device records, each associated with an audience record. The method further includes determining multiple keywords, each associated with an audience record and determining a position for each keyword in the embedding space. The method further includes calculating a first distance between the position of the search query in the embedding space and the position of each audience record in the embedding space. The method further includes calculating a second distance between the position of the search query in the embedding space and the position of each keyword in the embedding space. The method further includes ranking each audience record based on the first distance and the second distance.