This repository contains a PyTorch implementation of BERT to extract the portion of a tweet that represents it's sentiment. The code makes use of the huggingface library for pretrained versions of the BERT model.
sentiment-bert.ipynb contains the implementation and results. The data and problem description can be found in the Kaggle competition Tweet Sentiment Extraction.
For a detailed walk-through of this code, please view my blog post at BERT: Using Hugging Face for Sentiment Extraction with PyTorch.