Sentiment Analysis of English-Hindi Code-Mixed Text Using mBERT Model

In this work, a multilingual BERT (mBERT) model for sentiment analysis of text with mixed English and Hindi codes is presented. A common feature of casual communication, especially on digital platforms, code-mixing poses serious challenges due to its inconsistent grammar and jumbled vocabulary. The suggested method addresses this by enhancing mBERT to categorize sentiments into five groups that encompass a broad range of emotions. Careful planning ensures uniform input formatting, and the dataset contains multilingual user inputs from real-world scenarios. Common measures like F1-score, recall, accuracy, and precision are used to assess the model. The findings show that mBERT can handle code-mixed input efficiently without the need for languagespecific modifications. This solution proves efficient in multilingual sentiment classification tasks and sets the groundwork for future enhancements in handling low-resource and mixed-language data. The study highlights the relevance of transformer-based models in real-time multilingual analysis and offers practical insights for expanding sentiment analysis in linguistically diverse environments.