The code in this repository features sentiment analysis with VADER, TextBlob and tranfer learning wth BERT + trained regression model. The sentiment is positive, negative and neutral. The BERT-based model is binary,positive/negative. For more details regarding methods, result and models see the paper "Sentiment analysis on Twitter data towards climate action", Emelie Rosenberg, Carlota Tarazona, Fermín Mallor, Hamidreza Eivazi, David Pastor-Escuredo, Francesco Fuso-Nerini, Ricardo Vinuesa
This site was built using GitHub Pages.
This repository is transferred from the public repository. The transfer of this repository is conducted in accordance with the permission granted by the original author. For inquiries regarding this repository, please contact author.
- Training: The training of the Regression model was done with the Kaggle dataset Sentiment140.
- Collected data: It was collected with the python package, Tweepy. For the collecte draw data see Google drive file.
- Annotated data: 247 tweets was annoteded as positive, neutral and negative.
The TweetBert model was used for tokanization. For more information about the model see Hugging Face model card. The BERT-model was runned on GPU.
The model can be found under code/Sentiment analysis/Model/. The regression model was trained on Google Colab.