Predicting Hate Intensity of Twitter Conversation Threads

Abstract

Tweets are the most concise form of communication in online social media, wherein a single tweet has the potential to make or break the discourse of the conversation. Online hate speech is more accessible than ever, and stifling its propagation is of utmost importance for social media companies and users for congenial communication. Most of the research barring a recent few has focused on classifying an individual tweet regardless of the tweet thread/context leading up to that point. One of the classical approaches to curb hate speech is to adopt a reactive strategy after the hate speech postage. The ex-post facto strategy results in neglecting subtle posts that do not show the potential to instigate hate speech on their own but may portend in the subsequent discussion ensuing in the post's replies.In this paper, we propose DRAGNET++, which aims to predict the intensity of hatred that a tweet can bring in through its reply chain in the future. It uses the semantic and propagating structure of the tweet threads to maximize the contextual information leading up to and the fall of hate intensity at each subsequent tweet. We explore three publicly available Twitter datasets -Anti-Racism contains the reply tweets of a collection of social media discourse on racist remarks during US political and Covid-19 background;Anti-Social presents a dataset of 40 million tweets amidst the COVID-19 pandemic on anti-social behaviours with custom annotations; and Anti-Asian presents Twitter datasets collated based on anti-Asian behaviours during COVID-19 pandemic. All the curated datasets consist of structural graph information of the Tweet threads. We show that DRAGNET++ outperforms all the state-of-the-art baselines significantly. It beats the best baseline by an 11% margin on the Person correlation coefficient and a decrease of 25% on RMSE for the Anti-Racism dataset with a similar performance on the other two datasets.

On the Identification and Forecasting of Hate Speech in Inceldom

“…In future work, we plan to delve further into forecasting by implementing temporal and propagation features (e.g., Meng et al (2023); Dahiya et al (2021); Lin et al (2021); Almerekhi et al (2020); Jaki et al (2019)). Based on Pelicon et al (2021), we also plan to expand language coverage, with German- (Mandl et al, 2019) and Spanishlanguage (Basile et al, 2019) hate speech datasets being two of the most prominent candidates due to their similarity to English and Italian, respectively.…”

“…Zhang et al (2018) extract politeness strategies and rhetorical prompts to predict whether a conversation will turn uncivil. Meng et al (2023) predict the intensity of hate that a tweet might carry through its reply chain by exploiting tweet threads and their semantic and propagating structures. Dahiya et al (2021), compiled a dataset of 4.5k tweets and their reply threads, confirming that longitudinal patterns of hate intensity among reply threads are diverse, with no significant correlation with the source tweet.…”

Model-Agnostic Meta-Learning for Multilingual Hate Speech Detection

“…The classification heads are implemented using the pytorch 4 linear layers, initialized randomly. For implementing the meta-learning features (for example, first-order approximation), we use the learn2learn library 5 .…”

“…The gravity of the situation has motivated social media platforms and academic researchers to propose traditional machine learning and deep learning solutions to detect online hate speech automatically [1]- [3] and early [4], [5]. Among the solutions, large pre-trained language models (LMs), which have demonstrated their superiority in many NLP tasks [6],…”