Deep-Learning-Projects

Hotel-Reviews-Classifier

1. Classified hotel reviews into 2-class classification as positive/negative and deceptive/truthful reviews using 2-layered network using TensorFlow.
2. Implemented our own gradient descent and even visualize test data by reducing dimensions using TSNE. You can see in tsne_embeds.pdf
3. Training data used is same as under Naive Bayes Classifier under NLP Project Repository.
4. Run starter.py with input data as command line argument.
5. Got 95% accuracy.

RNN-POS-Tagger

1. Tagged Italian, Japanese and surprise language using sequence to sequence model using TensorFlow.
2. Used Bidirectional LSTM to get better accuracy.
3. Got 96% accuracy for Japanese and Italian languages.