Text Classification Models with Tensorflow

Tensorflow implementation of Text Classification Models.

Implemented Models:

1. Word-level CNN [paper]
2. Character-level CNN [paper]
3. Very Deep CNN [paper]
4. Word-level Bidirectional RNN
5. Attention-Based Bidirectional RNN [paper]
6. RCNN [paper]

Semi-supervised text classification(Transfer learning) models are implemented at [dongjun-Lee/transfer-learning-text-tf].

Requirements

• Python3
• Tensorflow
• pip install -r requirements.txt

Usage

Train

To train classification models for dbpedia dataset,

$ python train.py --model="<MODEL>"

(<Model>: word_cnn | char_cnn | vd_cnn | word_rnn | att_rnn | rcnn)

Test

To test classification accuracy for test data after training,

$ python test.py --model="<TRAINED_MODEL>"

Sample Test Results

Trained and tested with dbpedia dataset. (dbpedia_csv/train.csv, dbpedia_csv/test.csv)

(SA-LSTM and LM-LSTM are implemented at [dongjun-Lee/transfer-learning-text-tf].)

Models

1. Word-level CNN

Implementation of Convolutional Neural Networks for Sentence Classification.

2. Char-level CNN

Implementation of Character-level Convolutional Networks for Text Classification.

3. Very Deep CNN (VDCNN)

Implementation of Very Deep Convolutional Networks for Text Classification.

4. Word-level Bi-RNN

Bi-directional RNN for Text Classification.

1. Embedding layer
2. Bidirectional RNN layer
3. Concat all the outputs from RNN layer
4. Fully-connected layer

5. Attention-Based Bi-RNN

Implementation of Attention-Based Bidirectional Long Short-Term Memory Networks for Relation Classification.

6. RCNN

Implementation of Recurrent Convolutional Neural Networks for Text Classification.

References

• dennybritz/cnn-text-classification-tf
• zonetrooper32/VDCNN