Sentiment classification is a popular task in NLP. This notebook uses wor2vec representations and compares various classifiers: from the typical k-NN to the more advanced LSTM networks. For LSTMs specifically, we are interested in using both pre-trained and not pre-trained embeddings. The code is based on the following works of:
https://www.kaggle.com/ngyptr/lstm-sentiment-analysis-keras/data
https://www.kaggle.com/lystdo/lstm-with-word2vec-embeddings
Other sources:
https://www.kaggle.com/pierremegret/gensim-word2vec-tutorial
https://keras.io/getting-started/sequential-model-guide/
https://adventuresinmachinelearning.com/word2vec-keras-tutorial/