Like a Good Nearest Neighbor: Practical Content Moderation and Text Classification

TL;DR

This paper introduces LaGoNN, a parameter-free modification to SetFit that enhances text classification and content moderation by incorporating nearest neighbor information, improving performance especially in multilingual and domain drift scenarios.

Contribution

LaGoNN is a novel, parameter-free method that modifies input texts with neighbor information to improve text classification and content moderation.

Findings

LaGoNN improves SetFit's performance in content moderation tasks.

LaGoNN effectively flags undesirable content across multiple languages.

The method enhances classification robustness under domain drift.

Abstract

Few-shot text classification systems have impressive capabilities but are infeasible to deploy and use reliably due to their dependence on prompting and billion-parameter language models. SetFit (Tunstall et al., 2022) is a recent, practical approach that fine-tunes a Sentence Transformer under a contrastive learning paradigm and achieves similar results to more unwieldy systems. Inexpensive text classification is important for addressing the problem of domain drift in all classification tasks, and especially in detecting harmful content, which plagues social media platforms. Here, we propose Like a Good Nearest Neighbor (LaGoNN), a modification to SetFit that introduces no learnable parameters but alters input text with information from its nearest neighbor, for example, the label and text, in the training data, making novel data appear similar to an instance on which the model was optimized. LaGoNN is effective at flagging undesirable content and text classification, and improves the performance of SetFit. To demonstrate the value of LaGoNN, we conduct a thorough study of text classification systems in the context of content moderation under four label distributions, and in general and multilingual classification settings.