Hitachi at SemEval-2020 Task 12: Offensive Language Identification with Noisy Labels using Statistical Sampling and Post-Processing

TL;DR

This paper describes a hybrid BERT-based system for offensive language detection in noisy Twitter data, utilizing statistical sampling and post-processing, achieving high F1 scores and providing insights for future research.

Contribution

The paper introduces a novel combination of statistical sampling and post-processing techniques with BERT for noisy label offensive language detection.

Findings

Achieved Macro-F1 score of 0.90913 on SemEval-2020 Task-12 dataset.

Developed a hybrid system combining sampling, BERT, and post-processing.

Provided detailed error analysis to guide future research.

Abstract

In this paper, we present our participation in SemEval-2020 Task-12 Subtask-A (English Language) which focuses on offensive language identification from noisy labels. To this end, we developed a hybrid system with the BERT classifier trained with tweets selected using Statistical Sampling Algorithm (SA) and Post-Processed (PP) using an offensive wordlist. Our developed system achieved 34 th position with Macro-averaged F1-score (Macro-F1) of 0.90913 over both offensive and non-offensive classes. We further show comprehensive results and error analysis to assist future research in offensive language identification with noisy labels.