char-conv-classification

Character level convolutional neural network for text classification

Link to the paper

Character-level Convolutional Networks for Text Classification

Network Architecture

This architecture deviates slightly from the network specified in the paper. It started as a character conv, but soon it moved to word2vec, because of the lack of powerful system. It is 9 layer convolutional neural network. Convolution layers has number of filters equal to 256. Number of units at the output are 4 (depends on the dataset). Here is the architecture table taken from the paper.

Results