Incremental user embedding modeling for personalized text classification
This work addresses the problem of personalized text classification for applications like chatbots and social media, but it is incremental as it builds on existing user embedding methods with a dynamic integration approach.
The paper tackled the challenge of adaptive user representation learning from growing interaction histories by proposing an incremental user embedding modeling approach that integrates recent interactions into accumulated history vectors using a transformer encoder, achieving 9% and 30% relative improvements in prediction accuracy over a baseline on a personalized multi-class classification task with the Reddit dataset.
Individual user profiles and interaction histories play a significant role in providing customized experiences in real-world applications such as chatbots, social media, retail, and education. Adaptive user representation learning by utilizing user personalized information has become increasingly challenging due to ever-growing history data. In this work, we propose an incremental user embedding modeling approach, in which embeddings of user's recent interaction histories are dynamically integrated into the accumulated history vectors via a transformer encoder. This modeling paradigm allows us to create generalized user representations in a consecutive manner and also alleviate the challenges of data management. We demonstrate the effectiveness of this approach by applying it to a personalized multi-class classification task based on the Reddit dataset, and achieve 9% and 30% relative improvement on prediction accuracy over a baseline system for two experiment settings through appropriate comment history encoding and task modeling.