This is slightly simplified implementation of sentence intent classifer. Mainly refer to Convolutional Neural Networks for Sentence Classification paper

Requirements

• Python 3.6
• Tensorflow 1.4.0
• Numpy

Training

Print parameters:

python3.6 cnn_cls.py --help

optional arguments:
  -h, --help           show this help message and exit
  --kernel KERNEL      stem_cnn:static-embedding cnn, nonstem_cnn:non static-
                       embedding cnn, multiem_cnn:static and non static
                       embedding cnn, svm:linear svm
  --train [TRAIN]      if True, begin to train
  --data_dir DATA_DIR  Directory containing corpus

Train:

python3.6 cnn_cls.py --Train True --kernel stem_cnn
using static word embedding which learned by word2vec as the input of cnn

python3.6 cnn_cls.py --Train True --kernel nonstem_cnn
using non static word embedding which learned by model itself as the input of cnn

python3.6 cnn_cls.py --Train True --kernel multiem_cnn
using non static word embedding and static word embedding as the input of cnn

python3.6 cnn_cls.py --Train True --kernel svm
using linear svm

Evaluating

python3.6 cnn_cls.py --Train False --kernel stem_cnn

python3.6 cnn_cls.py --Train False --kernel nonstem_cnn

python3.6 cnn_cls.py --Train False --kernel multiem_cnn

python3.6 cnn_cls.py --Train False --kernel svm

Defalut corpus dictory is atis, you can use --data_dir to use your own data

References

• Convolutional Neural Networks for Sentence Classification