Personalization Module Combining Encoder Pretrained Learned Representations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing online marketplace systems lack effective methods to provide personalized and relevant advertisements to users based on real-time user actions, leading to suboptimal engagement and conversion rates.
Innovation Solution
A computer-implemented method using a diversifiable personalization module that generates short-term user representations by combining encoder, pretrained, and learned representations of user actions within a sliding window, enabling personalized ranking and prediction models for click-through rates and post-click conversion rates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional non-personalized ranking and bidding systems are used, then system complexity is low, but advertisement relevance and user engagement are insufficient
Solution Approach 1:
The system segments user behavior analysis into multiple independent representation components: encoder representations, pretrained representations, and learned representations. Each component processes user actions independently within a sliding window, then their outputs are combined to form comprehensive user profiles. This segmentation allows the system to achieve high advertisement relevance through multiple specialized modules while managing complexity by keeping each module focused and independent.
Solution Approach 2:
The personalization module is designed as a universal component that can be applied across different downstream tasks including click-through rate prediction, conversion rate prediction, and advertisement ranking. The same module that generates user representations from user actions can be reused for multiple prediction objectives, reducing overall system complexity while maintaining high adaptability across different advertisement personalization scenarios.
2Adaptability or versatility
If real-time user action analysis is implemented, then advertisement personalization improves, but computational resources and processing time increase
Solution Approach 1:
The system uses a sliding window approach that processes only a subset of user actions within a specific time window, rather than analyzing the entire user history. This partial action approach focuses computational resources on the most recent and relevant user behaviors, achieving effective personalization while significantly reducing the computational burden compared to processing complete user action histories.
Solution Approach 2:
The system pre-processes user actions by generating multiple types of representations (encoder, pretrained, and learned representations) that are stored and ready for combination. This preliminary action allows the personalization module to quickly generate user profiles when needed for advertisement ranking, reducing real-time computational requirements while maintaining high personalization accuracy.
3Measurement precision
If multiple representation types are combined, then prediction accuracy improves, but model complexity and training difficulty increase
Solution Approach 1:
The model segments the representation learning process into three distinct components: encoder representations that capture sequential patterns, pretrained representations that provide domain knowledge, and learned representations that adapt to specific tasks. Each component is trained independently on its own objectives, then their outputs are combined through simple concatenation or addition. This segmentation allows the system to achieve high prediction accuracy by leveraging diverse representation types while managing model complexity through independent training and simple combination operations.
Data Source
AI summary
Aspects of the disclosure provide a computer-implemented method for generating personalized results. The method includes identifying a set of user actions by a specific user within a sliding window of time, generating a first representations for the set of user actions using an encoder component of a personalization module, generating a second representation for the set of user actions using a pretrained representations component of the personalization module, generating a third representation for the set of user actions using a learned representations component of the personalization module, using the personalization module to combine the first representation, second representation and the third representation to generate a short-term personalized representation for the specific user, and providing a set of results for display to the user based on the short-term personalized representation.