NLPDove at SemEval-2020 Task 12: Improving Offensive Language Detection with Cross-lingual Transfer

TL;DR

This paper enhances multilingual offensive language detection by employing semi-supervised data, cross-lingual transfer with data selection, and specialized preprocessing, leading to improved performance across several languages.

Contribution

It introduces a new transferability metric, Translation Embedding Distance, and demonstrates effective cross-lingual transfer and data augmentation strategies for offensive language detection.

Findings

Performance improved with semi-supervised data

Cross-lingual transfer with data selection is effective

Achieved competitive results in Greek, Danish, and Turkish

Abstract

This paper describes our approach to the task of identifying offensive languages in a multilingual setting. We investigate two data augmentation strategies: using additional semi-supervised labels with different thresholds and cross-lingual transfer with data selection. Leveraging the semi-supervised dataset resulted in performance improvements compared to the baseline trained solely with the manually-annotated dataset. We propose a new metric, Translation Embedding Distance, to measure the transferability of instances for cross-lingual data selection. We also introduce various preprocessing steps tailored for social media text along with methods to fine-tune the pre-trained multilingual BERT (mBERT) for offensive language identification. Our multilingual systems achieved competitive results in Greek, Danish, and Turkish at OffensEval 2020.