Covering Cracks in Content Moderation: Delexicalized Distant Supervision for Illicit Drug Jargon Detection

TL;DR

This work tackles illicit drug jargon moderation in user-generated content, where traditional banlists fail against evasive terms and euphemisms. It introduces JEDIS, a context-based detector trained with delexicalized distant supervision that combines a context-only module and a word-attribute module through an ensemble, enabling robust detection of unseen terms. The authors construct two evaluation datasets from Reddit and Silk Road Forum content, demonstrate that JEDIS yields higher F1 and substantially broader jargon coverage than word-based baselines, and provide qualitative and ablation analyses to illustrate its robustness to euphemisms and noisy negatives. The approach offers a scalable, domain-adaptive solution for practical moderation and demonstrates potential for cross-domain transfer and adaptation to other forbidden-content domains.

Abstract

In light of rising drug-related concerns and the increasing role of social media, sales and discussions of illicit drugs have become commonplace online. Social media platforms hosting user-generated content must therefore perform content moderation, which is a difficult task due to the vast amount of jargon used in drug discussions. Previous works on drug jargon detection were limited to extracting a list of terms, but these approaches have fundamental problems in practical application. First, they are trivially evaded using word substitutions. Second, they cannot distinguish whether euphemistic terms such as "pot" or "crack" are being used as drugs or in their benign meanings. We argue that drug content moderation should be done using contexts rather than relying on a banlist. However, manually annotated datasets for training such a task are not only expensive but also prone to becoming obsolete. We present JEDIS, a framework for detecting illicit drug jargon terms by analyzing their contexts. JEDIS utilizes a novel approach that combines distant supervision and delexicalization, which allows JEDIS to be trained without human-labeled data while being robust to new terms and euphemisms. Experiments on two manually annotated datasets show JEDIS significantly outperforms state-of-the-art word-based baselines in terms of F1-score and detection coverage in drug jargon detection. We also conduct qualitative analysis that demonstrates JEDIS is robust against pitfalls faced by existing approaches.

Figures (5)

• Figure 1: Content moderation procedures of word-based system and context-based system. Bolded words in Input were predicted by each system. In the Result, words detected as drug jargon are colored red, and words predicted as non-drugs are colored blue.
• Figure 2: Overview of JEDIS.
• Figure 3: Jargon detection performance with varying values of $\mathcal{R}_{nonSTMS}$. $\mathcal{R}_{STMS}$ is set to 5.
• Figure 4: Jargon detection performance with varying contextual window sizes.
• Figure 5: Prompts fed into GPT4o-mini. The definition of drug jargon in Section \ref{['preliminaries']} and the annotation guidelines in Appendix \ref{['appendix']} are used.