Overview of the HASOC Subtrack at FIRE 2023: Identification of Conversational Hate-Speech

Madhu, H and Satapara, S and Pandya, P and Shah, N and Mandl, T and Modha, S (2023) Overview of the HASOC Subtrack at FIRE 2023: Identification of Conversational Hate-Speech. In: 15th Forum for Information Retrieval Evaluation, FIRE 2023, 15 December 2023through 18 December 2023, Goa,India, pp. 351-359.

PDF CEUR_wor_pro_3681_2023.pdf - Published Version Restricted to Registered users only Download (243kB) | Request a copy

Abstract

Identifying hate speech based on context is a requirement for real-world content moderation systems. However, in research, the definition and use of context for hate speech recognition has seen a variety of approaches. The task�Identification of Conversational Hate Speech� 2023 has provided a further dataset for hate speech detection, including context. The data was collected from Twitter (called X since 2023) and includes tweets, comments, or responses to such tweets or comments. This paper reports on the dataset, experiments, and results. Six teams submitted results for the binary classification task, and the best submission reached an F1 measure of 0.8. For the second task, five submissions were submitted. We also present a baseline that uses unlabelled data to obtain its predictions. © 2023 Copyright for this paper by its authors.

Item Type:	Conference Paper
Publication:	CEUR Workshop Proceedings
Publisher:	CEUR-WS
Additional Information:	The copyright for this article belongs to ACM SIGIR Special Interest Group on Information Retrieval.
Keywords:	Classification (of information); Deep learning; Social networking (online); Text processing, Benchmark; Context; Deep learning; Evaluation; Hate speech; Language resources; Real-world; Social media; Task Identification; Text classification, Speech recognition
Department/Centre:	Division of Electrical Sciences > Electrical Communication Engineering
Date Deposited:	24 Sep 2024 06:56
Last Modified:	24 Sep 2024 06:56
URI:	http://eprints.iisc.ac.in/id/eprint/85224

Actions (login required)

View Item