Sentiment analysis for a highly imbalanced financial news dataset using Tensorflow and RoBERTa tokenizer and embeddings.
Some custom layers are added for classification task.
Main purpose of this project was to get started with a text classification task in Tensorflow with RoBERTa tokenizer and embeddings of Transformers from 🤗 and fine-tuning.
Dataset credit goes to: https://www.kaggle.com/ankurzing
Dataset Link: https://www.kaggle.com/ankurzing/sentiment-analysis-for-financial-news