Generalisability of Topic Models in Cross-corpora Abusive Language Detection

Tulika Bose, Irina Illina, Dominique Fohr


Abstract
Rapidly changing social media content calls for robust and generalisable abuse detection models. However, the state-of-the-art supervised models display degraded performance when they are evaluated on abusive comments that differ from the training corpus. We investigate if the performance of supervised models for cross-corpora abuse detection can be improved by incorporating additional information from topic models, as the latter can infer the latent topic mixtures from unseen samples. In particular, we combine topical information with representations from a model tuned for classifying abusive comments. Our performance analysis reveals that topic models are able to capture abuse-related topics that can transfer across corpora, and result in improved generalisability.
Anthology ID:
2021.nlp4if-1.8
Volume:
Proceedings of the Fourth Workshop on NLP for Internet Freedom: Censorship, Disinformation, and Propaganda
Month:
June
Year:
2021
Address:
Online
Editors:
Anna Feldman, Giovanni Da San Martino, Chris Leberknight, Preslav Nakov
Venue:
NLP4IF
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
51–56
Language:
URL:
https://aclanthology.org/2021.nlp4if-1.8
DOI:
10.18653/v1/2021.nlp4if-1.8
Bibkey:
Cite (ACL):
Tulika Bose, Irina Illina, and Dominique Fohr. 2021. Generalisability of Topic Models in Cross-corpora Abusive Language Detection. In Proceedings of the Fourth Workshop on NLP for Internet Freedom: Censorship, Disinformation, and Propaganda, pages 51–56, Online. Association for Computational Linguistics.
Cite (Informal):
Generalisability of Topic Models in Cross-corpora Abusive Language Detection (Bose et al., NLP4IF 2021)
Copy Citation:
PDF:
https://aclanthology.org/2021.nlp4if-1.8.pdf