/tamil_mix_sentiment_analysis

NLP sentiment analysis of Tamil and English mixed text

Primary LanguageJupyter NotebookMIT LicenseMIT

NLP sentiment analysis of Tamil and English mixed text

In his notebook I create a fine-tuned model for sentiment analysis of text written in mixed a mixture of Tamil and English language.

I use a weighted version of the loss function to account for the great imbalance in the classes of the dataset. I also tune the tokenizer to recognize emojis as those are generally relevant to express sentiment. My model is based on a pre-trained multilingual model which includes Tamil and English; this should help in the text classification process.

The dataset was firs published in the UCI Machine Learning Repository and later published in the Hugging Face Data Repository from where I read it in this notebook. The original paper explaining the data set can be found at: https://aclanthology.org/2020.sltu-1.28.pdf

Links