Compare BERT-based models for document-level sentiment analysis using the SemEval 2017 Twitter dataset.
Run the following command to install other dependencies
pip install -r requirements.txtDownload the SemEval twitter data from here and place the data from the Subtask A from the GOLD folder into the data/tweets folder.
The main program can be run by the following command:
python main.py trainCompare different preprocessing models using the --preprocess_model flag. Choose between tfidf, bert-base-uncased, or other models from Huggingface. The program compares different 'head' neural models using the generated embeddings.
More options can be seen by using the -h tag.