Non-Parametric Temporal Adaptation for Social Media Topic Classification

TL;DR

This paper introduces a non-parametric dense retrieval method for social media topic classification that adapts to temporal changes and data deletion without re-training, outperforming traditional models on a year-long Twitter dataset.

Contribution

The paper proposes a novel non-parametric dense retrieval technique for temporal adaptation in social media NLP tasks, eliminating the need for re-training and handling data deletion efficiently.

Findings

Improves hashtag prediction accuracy by 64.12% over parametric baselines.

Effectively adapts to temporal distribution shifts in social media data.

Handles user data deletion with negligible performance loss.

Abstract

User-generated social media data is constantly changing as new trends influence online discussion and personal information is deleted due to privacy concerns. However, most current NLP models are static and rely on fixed training data, which means they are unable to adapt to temporal change -- both test distribution shift and deleted training data -- without frequent, costly re-training. In this paper, we study temporal adaptation through the task of longitudinal hashtag prediction and propose a non-parametric dense retrieval technique, which does not require re-training, as a simple but effective solution. In experiments on a newly collected, publicly available, year-long Twitter dataset exhibiting temporal distribution shift, our method improves by 64.12% over the best parametric baseline without any of its costly gradient-based updating. Our dense retrieval approach is also particularly well-suited to dynamically deleted user data in line with data privacy laws, with negligible computational cost and performance loss.