Reddit hate narratives isolated by clustering entity based evaluations
Look What You Made Us Cluster: Hate Narrative Extraction from Reddit Discourse
Computation and LanguageSocial and Information Networks
Summary
The authors focus on detecting hate narratives on Reddit more precisely by looking at not just words but who or what is being talked about and how they are judged. They created a process that uses a large AI language model to identify pairs of entities and evaluations, then groups similar narratives together to better understand hate speech patterns. This method was demonstrated with posts criticizing Taylor Swift, showing how it can reveal clear hate speech narratives. This approach aims to provide clearer, more interpretable analysis of online hate than previous methods that only looked at surface meaning.
What this means in practice
- •For online content moderators: Identify and group hate speech narratives with clearer context for improved moderation decisions on social platforms.
- •For social media analysis teams: Analyze clustered hate narratives to understand community sentiment and discourse dynamics related to public figures.
Tested on one dataset.
Authors
Annabelle K. L. Chua, Forster J. Khoo, Joel C. R. Tan, Huey Ting Ang, Kheng Hwee Tan, Joel Y. A. Sim, Shirley W. H. Ow, Ria Mundhra, Elsie C. K. Toh, Youfeng Xu, Lynnette H. X. Ng
Abstract
Narrative extraction allows us to identify online hate narratives, supporting the construction of rigorous detection systems. Existing computational approaches, however, are limited in precision as they rely on semantic representations, which tend to capture only surface-level meaning. To detect more precise and interpretable narratives, we present an extraction pipeline that represents narratives as entity-evaluation pairs. Narratives are extracted using a Large Language Model (LLM) reasoning process that extends Aspect-Based Sentiment Analysis, identifying the aspect, classifying its judgement type as the basis for evaluation, and deriving the evaluation accordingly. Extracted narratives are then clustered using Leiden, following which clusters are resolved to an intended level of granularity through an LLM-guided refinement process. We illustrate this narrative pipeline with English Reddit comments from 2024 that criticize Taylor Swift, analyzing a representative cluster that exhibits hate speech patterns to demonstrate its interpretive value.