Text Classification

• Dataset: yahoo/dbpedia...
• Model: CNN/(Bi)LSTM/(Bi)GRU.
• Pytorch: 1.0.1
• Python: 3.6
• torchtext: 0.3.1.
• Support pretrained word embedding.(word2vec,glove)

torchtext

• This package can provide an elegant way to build vocabulary(torchtext).

TEXT.build_vocab(dataset, vectors)

Training

python preprocess.py #preprocessing

python main.py #training

Refs

• Kim, Y. J. E. A. (2014). "Convolutional Neural Networks for Sentence Classification."
• cnn-text-classification-pytorch
• Understanding LSTM Networks
• Glove
• 你在训练RNN的时候有哪些特殊的trick？
• CNN
• rnn.py
• rnn initialization
• Explaining and illustrating orthogonal initialization for recurrent neural networks