Plan of Action
- Load Training.1600000.processed.noemoticon.csv Dataset (1.6 Mn twitter tweets) and a Twitter Finance Tweets Datasets
- Pre-process dataset by removing special characters, numbers, etc. from user reviews + convert sentiment labels positive & negative to numbers 1 & 0, respectively
- Import GloVe Word Embedding to build Embedding Dictionary + Use this to build Embedding Matrix for our Corpus
- Model Training using Deep Learning in Keras for separate: Simple Neural Net, CNN, LSTM and DistilBERT Models and analyse model performance and results
- Last, perform predictions on real L&T tweets