TalkDown: A Corpus for Condescension Detection in Context

TL;DR

This paper introduces TalkDown, a new dataset for detecting condescension in context, and demonstrates that incorporating discourse information improves detection accuracy, enabling analysis of community norms across online platforms.

Contribution

The paper presents TalkDown, the first labeled dataset for condescension detection in context, and proposes models that leverage discourse representations to improve detection performance.

Findings

Discourse-aware models outperform language-only models in condescension detection.

Condescension rates vary significantly across online communities.

Techniques for handling low condescension prevalence improve detection robustness.

Abstract

Condescending language use is caustic; it can bring dialogues to an end and bifurcate communities. Thus, systems for condescension detection could have a large positive impact. A challenge here is that condescension is often impossible to detect from isolated utterances, as it depends on the discourse and social context. To address this, we present TalkDown, a new labeled dataset of condescending linguistic acts in context. We show that extending a language-only model with representations of the discourse improves performance, and we motivate techniques for dealing with the low rates of condescension overall. We also use our model to estimate condescension rates in various online communities and relate these differences to differing community norms.