This is an implementation of `CharSCNN`, and is heavily based on (Theano's CNN)[http://www.deeplearning.net/tutorial/lenet.html#lenet] This code is an implementation exercise of CharSCNN, and might not reflect original ideas. We achieved about 70% accuracy, on a different dataset than original paper, and a network with smaller dimensions thanoriginal one. The network was trained on a powerful non GPU machine. Read more in Report.pdf to know about other details.