Reed at SemEval-2020 Task 9: Fine-Tuning and Bag-of-Words Approaches to Code-Mixed Sentiment Analysis

TL;DR

This paper presents approaches for sentiment analysis on Hinglish tweets using fine-tuned BERT models and bag-of-words neural networks, achieving competitive results in the SemEval-2020 SentiMix task.

Contribution

It introduces the application of transfer learning and bag-of-words methods to code-mixed sentiment analysis, demonstrating effective strategies for Hinglish data.

Findings

Achieved an F-score of 71.3% with the best model.

Placed 4th out of 62 entries in the competition.

Showed the effectiveness of transfer learning for code-mixed sentiment analysis.

Abstract

We explore the task of sentiment analysis on Hinglish (code-mixed Hindi-English) tweets as participants of Task 9 of the SemEval-2020 competition, known as the SentiMix task. We had two main approaches: 1) applying transfer learning by fine-tuning pre-trained BERT models and 2) training feedforward neural networks on bag-of-words representations. During the evaluation phase of the competition, we obtained an F-score of 71.3% with our best model, which placed $4^{th}$ out of 62 entries in the official system rankings.