Overview of the HASOC track at FIRE 2019: Hate Speech and Offensive Content Identification in Indo-European Languages
By Thomas Mandl, Sandip J Modha, Prasenjit Majumder, Daksh Patel, Mohana Dave, Chintak Mandalia, Aditya Patel. 2019
407 citations recorded by the source corpus.
The identification of Hate Speech in Social Media is of great importance and receives much attention in the text classification community. There is a huge demand for research for languages other than English. The HASOC track intends to stimulate development in Hate Speech for Hindi, German and English. Three datasets were developed from Twitter and Facebook and made available. Binary classification and more fine-grained subclasses were offered in 3 subtasks. For all subtasks, 321 experiments were submitted. The approaches used most often were LSTM networks processing word embedding input. The performance of the best system for identification of Hate Speech for English, Hindi, and German was a Marco-F1 score of 0.78, 0.81 and 0.61, respectively.
Subfields: Artificial Intelligence, Information Extraction, AI Ethics, Fairness & Society
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Bertie Vidgen
- Valerio Basile
- Viviana Patti
- Douwe Kiela
- Endang Wahyu Pamungkas
- Preslav Nakov
- Tristan Thrush
- Leon Derczynski
In Information Extraction
In AI Ethics, Fairness & Society
- Bertie Vidgen
- Marcos Zampieri
- Valerio Basile
- Leon Derczynski
- Aida Mostafazadeh Davani
- Viviana Patti
- Helen Z. Margetts
- Dong Nguyen
Citations in the corpus
- Learning from the Worst: Dynamically Generated Datasets to Improve Online Hate Detection (2021, background): “Benchmark datasets Several benchmark datasets have been put forward for online hate classification (Waseem and Hovy, 2016; Waseem, 2016; Davidson et al., 2017; Founta et al., 2018; Mandl et al., 2019; Zampieri et al., 2019, 2020; Vidgen et al., 2020, 2021).”
- SemEval-2020 Task 12: Multilingual Offensive Language Identification in Social Media (OffensEval 2020) (2020, methodology): “Examples include HASOC 2019 (Mandl et al., 2019) for English, German, and Hindi, HatEval 2019 (Basile et al., 2019) for English and Spanish, GermEval 2019 for German (Struß et al., 2019), and TRAC 2020 (Kumar et al., 2018b) for Bengali, Hindi, and English.”
- Resources and benchmark corpora for hate speech detection: a systematic review (2020, methodology): “Finally, two competitions explicitly focused on the identification of both HS and offensive language, i.e. HASOC at FIRE 2019 (Mandl et al. 2019) and HSD , the HS detection task on Vietnamese at VLSP campaign in 2019 (Vu et al. 2019).”
- Toxic Language Detection in Social Media for Brazilian Portuguese: New Dataset and Multilingual Analysis (2020, background): “…et al., 2018) –, most of the resources and studies available are for English (Davidson et al., 2017; Wulczyn et al., 2017; Founta et al., 2018; Mandl et al., 2019; Zampieri et al., 2019b).3 For Portuguese, only two previous works are available (Fortuna et al., 2019; de Pelle and Moreira,…”
- SOLID: A Large-Scale Semi-Supervised Dataset for Offensive Language Identification (2020, methodology): “The taxonomy was successfully adopted for several languages (Mubarak et al., 2021; Pitenis et al., 2020; Sigurbergsson and Derczynski, 2020; Çöl-tekin, 2020), and it was used in a series of shared tasks (Zampieri et al., 2019b; Mandl et al., 2019).”
- Introducing CAD: the Contextual Abuse Dataset (2021, background): “Recent work have addressed both types of content, recognising that they are conceptually different but often co-occur in the real-world and share syntactical and lexical similarities (Zampieri et al., 2019; Mandl et al., 2019).”
- Misogyny Detection in Twitter: a Multilingual and Cross-Domain Study (2020, positive): “Saha et al. (2019) employed Multilingual-BERT 4 and language agnostic sentence embedding representations (LASER Artetxe & Schwenk (2019)) as feature representation, trained on Light Gradient Boosting Machine (LGBM) Ke et al. (2017) to build a language-agnostic system to detect hate speech, in their ”
- Contextualizing Hate Speech Classifiers with Post-hoc Explanation (2020, methodology): “Specifically, we use post-hoc explanation algorithms to interpret and modulate fine-tuned language models like BERT (Devlin et al., 2018), which achieve state of the art performance on many hate speech detection tasks (MacAvaney et al., 2019; Mandl et al., 2019).”