A c-command Decoder: This project was done as a team project (with Jing Ji) for the NLP (CSE538) course taken at Stony Brook University. The goal of the project is to build a model that encodes c-command dependencies directly from strings rather than trees. The input for the model is parsed trees from Penn Treebank and the model was trained using Gated Recurrent Units and dense layers. This is still an ongoing project since the model seems to be suffering from overfitting due to the small size of data resulting in sparsity. Moreover, the use of dense layer seems to be discarding sequential information.
Sentiment Analysis: This project was done as a term project for AMS561 course at Stony Brook Univesity. This was a joint work with two other classmates. Using Python scikit-learn machine learning package, we did a sentiment analysis of Amazon reviews on Kindle e-books.