This is the code for paper Exploiting Topic-based Adversarial Neural Network for Cross-domain Keyphrase Extraction (ICDM 2018)
If you use the code, please kindly cite the paper:
@inproceedings{wang2018exploiting,
title={Exploiting Topic-based Adversarial Neural Network for Cross-domain Keyphrase Extraction},
author={Wang, Yanan and Liu, Qi and Qin, Chuan and Xu, Tong and Wang, Yijun and Chen, Enhong and Xiong, Hui},
booktitle={2018 IEEE International Conference on Data Mining (ICDM)},
pages={597--606},
year={2018},
organization={IEEE}
}
-
data
sampled data is provided in sample_data directory, you can prepare your own data including source domain and target domain data
vocab.pickle can be obtained using voc.py.
-
you can train the model using the following script, more parameters can be turned in adv_train.py
CUDA_VISIBLE_DEVICES=0 python adv_train.py \ --log_dir 'your log path' \ --CORPUS 'source_domain,target_domain' \ --batch_size 64 \ --dropout_keep_prob 0.5 \ --lstm_dim 300 \ --num_epochs 100 \ --num_filters 200 \ --word_dim 300 \ --evaluate_every 50 \ --conv_activation "elu" \ --early_stop 15 \ --use_gate 1 \ --lr 0.1 \ --clip 1 \ --lamda_type 3 \ --num_decode_steps 100 \ --lm_rate 0.2 \ --topic_num 50