Keras RNN model with SpaCy embedding for sentiment analysis of Kaggle challenge “Deep-NLP”:

https://www.kaggle.com/samdeeplearning/deepnlp

This code shows how SpaCy can help to get pretty decent results with sentiment classification with very limited data – only 125 resumes and 80 responses to a therapy chatbot.

I have used 80% data for testing and 20% for validation.

The model reached 100% accuracy on training data for both datasets. Validation accuracy is 84% for dataset of resumes (Sheet2.cvs) and 94% for responses to a chatbot (Sheet1.csv).