ExCode-Mixed: Explainable Approaches towards Sentiment Analysis on Code-Mixed Data using BERT models

TL;DR

This paper introduces explainable methods for sentiment analysis on code-mixed social media data using BERT models, addressing the need for interpretability in multilingual, mixed-language contexts.

Contribution

It presents a novel methodology to incorporate explainability techniques into BERT-based sentiment analysis for code-mixed data, enhancing interpretability.

Findings

Effective integration of explainability with BERT models for code-mixed sentiment analysis

Improved understanding of model predictions in multilingual contexts

Potential for better trust and usability of sentiment analysis tools

Abstract

The increasing use of social media sites in countries like India has given rise to large volumes of code-mixed data. Sentiment analysis of this data can provide integral insights into people's perspectives and opinions. Developing robust explainability techniques which explain why models make their predictions becomes essential. In this paper, we propose an adequate methodology to integrate explainable approaches into code-mixed sentiment analysis.