In this chapter, we have seen how to implement and deploy a hands-on deep learning project that classifies review texts as either positive or negative based on the words they contain. We have used a large-scale movie review dataset that contains 50,000 reviews (training plus testing). A combined approach using Word2Vec (that is, a widely used word embedding technique in NLP) and the LSTM network for modeling was applied: the pre-trained Google news vector model was used as the neural word embeddings.
Then, the training vectors, along with the labels, were fed into the LSTM network, which successfully classified them as negative or positive sentiments. Then, it evaluated the trained model on the test set. Additionally, we have also seen how to apply text-based preprocessing techniques such as tokenizer, stop words removal and TF-IDF, as well as word-embedding operations...
