Introduction:
We are building a deep learning model that predicts which tweets from Twitter are about real disasters and which ones aren’t. *This is a classification problem with NLP. *We came across this idea because it was a Kaggle competition and seemed like a cool thing to try out! *Our group will implement two architectures: LSTM and BERT for binary classification and compare the performance.
Challenges:
What has been the hardest part of the project you’ve encountered so far? *Translating idea to implementation without stencil code. Having to look up a lot of examples on how to use different methods and figuring out what would work best for our dataset. *Figuring out the sizes/dimensions of the tensors for LSTM. Fixed by print statements :) *Had trouble debugging LSTM because training loss was fluctuating but accuracy was relatively high (~77%). This was fixed by tuning hyperparameters.
Insights:
Are there any concrete results you can show at this point? How is your model performing compared with expectations? *We can show how we preprocess the dataset, the training and testing accuracy of BERT, and how we set up the LSTM model. *LSTM training loss is decreasing exponentially as expected and the test accuracy is around 75%. *BERT testing accuracy is 80%+.
Plan:
Are you on track with your project? *We still need to get training accuracies for both models and make graphs for visualization. Otherwise, we are on track with our project.
Log in or sign up for Devpost to join the conversation.