jsonko/tf-text-classification

CNN and RNN with Attention for Chinese Text Classification in Tensorflow

Environment setup

• Requirements: python3, tensorflow>=1.0, sklearn, numpy, pandas and jieba

pip install -r requirements.txt

Data preprocessing

• Text cleaning and word segmentation, data format: label\tsentence

python seg_words.py

Training models

• CNN

python train_cnn.py

• RNN with attention

python train_rnn.py

Evaluating models

python eval.py --checkpoint_dir=./runs/textrnn/trained_result_1512462690/checkpoints --model_type=RNN

If evaluation data has labels, has_label should be set to True.