Our project was to predict emojis associated with tweets. The dataset is taken from Kaggle:
https://www.kaggle.com/hariharasudhanas/twitter-emoji-prediction
The Stemmer is based on Apache OpenNLP:
The joinrows.py script is used to remove linebreaks in each tweet in the original dataset. To run the joinrows.py script:
python joinrows.py [in_file] [out_file]
The TweetStemmer is used to replace words in tweets by indexed stems (common roots of words). To run the TweetStemmer:
javac TweetStemmer.java
java TweetStemmer path/source.csv path/tweetStems.csv path/stems.csv
The main work including data analysis, Bayesian prediction and a neural network trained with Keras/TensorFlow is contained within twitter_emoji_prediction.ipynb.