Understanding writing style in social media with a supervised contrastively pre-trained transformer

TL;DR

This paper introduces STAR, a supervised contrastively pre-trained transformer model trained on a large social media corpus to improve authorship attribution and understanding of online harmful behaviors.

Contribution

We propose STAR, a novel author representation model trained on 4.5 million texts using supervised contrastive loss, achieving state-of-the-art zero-shot attribution and clustering performance.

Findings

Zero-shot attribution and clustering performance on PAN challenges

80% accuracy in author identification among 1616 authors

Effective authorship verification with a simple dense layer

Abstract

Online Social Networks serve as fertile ground for harmful behavior, ranging from hate speech to the dissemination of disinformation. Malicious actors now have unprecedented freedom to misbehave, leading to severe societal unrest and dire consequences, as exemplified by events such as the Capitol assault during the US presidential election and the Antivaxx movement during the COVID-19 pandemic. Understanding online language has become more pressing than ever. While existing works predominantly focus on content analysis, we aim to shift the focus towards understanding harmful behaviors by relating content to their respective authors. Numerous novel approaches attempt to learn the stylistic features of authors in texts, but many of these approaches are constrained by small datasets or sub-optimal training losses. To overcome these limitations, we introduce the Style Transformer for Authorship Representations (STAR), trained on a large corpus derived from public sources of 4.5 x 10^6 authored texts involving 70k heterogeneous authors. Our model leverages Supervised Contrastive Loss to teach the model to minimize the distance between texts authored by the same individual. This author pretext pre-training task yields competitive performance at zero-shot with PAN challenges on attribution and clustering. Additionally, we attain promising results on PAN verification challenges using a single dense layer, with our model serving as an embedding encoder. Finally, we present results from our test partition on Reddit. Using a support base of 8 documents of 512 tokens, we can discern authors from sets of up to 1616 authors with at least 80\% accuracy. We share our pre-trained model at huggingface (https://huggingface.co/AIDA-UPM/star) and our code is available at (https://github.com/jahuerta92/star)