This research revolves around tackling the issue of food insecurity by analyzing related tweets through Sentiment Analysis using Transfer and Deep Learning.
Using:
- Transfer Learning: IMDb movie reviews (training data) --> Food Insecurity tweets (testing data)
- Deep Learning: BiLSTM (Bidirectional Long Short Term Memory) Model
I mined 1558 tweets from Twitter using the Python Twitter API called tweepy, which I classifed the sentiment of (positive or negative). I used Deep Learning for this research, so I would need a lot of training data for the model that I used. I chose to use 50000 IMDb movie reviews as my training data.
I used a BiLSTM (Bidirectional Long Short Term Memory) model in order to classify the sentiment. The model's architecture is displayed in the below diagrams. Through research, unlike the standard LSTM approach, a BiLSTM model involves its inputs flowing in both directions and is thus capable of using information from both of its sides.
Using transfer learning, I trained this model with the IMDb data and tested it with the Food Insecurity Twitter data.