Gauravarora@HASOC-Dravidian-CodeMix-FIRE2020: Pre-training ULMFiT on Synthetically Generated Code-Mixed Data for Hate Speech Detection

TL;DR

This paper presents a method for hate speech detection in Dravidian languages using ULMFiT pre-trained on synthetically generated code-mixed data, achieving high F1-scores and ranking second in the competition.

Contribution

The novel approach involves pre-training ULMFiT on synthetically generated code-mixed data modeled as a Markov process, improving hate speech detection in Dravidian languages.

Findings

Achieved 0.88 F1-score for Tamil-English in Sub-task B

Ranked 2nd on the leaderboard for Tamil-English

Achieved 0.91 F1-score for Malayalam-English in Sub-task A

Abstract

This paper describes the system submitted to Dravidian-Codemix-HASOC2020: Hate Speech and Offensive Content Identification in Dravidian languages (Tamil-English and Malayalam-English). The task aims to identify offensive language in code-mixed dataset of comments/posts in Dravidian languages collected from social media. We participated in both Sub-task A, which aims to identify offensive content in mixed-script (mixture of Native and Roman script) and Sub-task B, which aims to identify offensive content in Roman script, for Dravidian languages. In order to address these tasks, we proposed pre-training ULMFiT on synthetically generated code-mixed data, generated by modelling code-mixed data generation as a Markov process using Markov chains. Our model achieved 0.88 weighted F1-score for code-mixed Tamil-English language in Sub-task B and got 2nd rank on the leader-board. Additionally, our model achieved 0.91 weighted F1-score (4th Rank) for mixed-script Malayalam-English in Sub-task A and 0.74 weighted F1-score (5th Rank) for code-mixed Malayalam-English language in Sub-task B.