Awesome

Improving Sequence Modeling Ability of Recurrent Neural Networks via Sememes

This repo is the official implementation of the paper "Improving Sequence Modeling Ability of Recurrent Neural Networks via Sememes" that is published in IEEE/ACM Transactions on Audio, Speech, and Language Processing (TASLP).

If you have any question about the code, feel free to contact: yujiaqin16@gmail.com & scouyang4354@gmail.com

Sememe Dataset

sememe.txt comprises all 2186 sememes. hownet_en.txt is an English version of HowNet sememe knowledge dictionary, where each line of word is followed by a line of all its sememes. More details about HowNet and sememes can be found in paper.

Language Modeling

First download PTB dataset and put it at data/ptb, then run:

cd LM
CUDA_VISIBLE_DEVICES=3 python main.py --cuda --emsize 1500 --nhid 1500  --epochs 40 --sememe_dim 1500 --model_type LSTM_cell --dropout 0.65

Natural Language Inference

First download the pretrained glove embeddings through: https://nlp.stanford.edu/projects/glove/ and put it at ../glove. Also download SNLI dataset through:

bash dataset/get_data.bash

and put it at ./dataset/SNLI.

Then train on SNLI Dataset by:

python3 train_nli.py --word_emb_path ../glove/glove.840B.300d.txt --encoder_type LSTM_cell --gpu_id 2

Text Classification

Save the encoder trained on SNLI at ./savedir, and test the downstream sentence enoding task using SentEval/encoders/infersent.py (you may need to modify the corresponding paths)

Adversarial Attack

For adversarial attack, you could generate adversarial examples after you have trainning an NLI model. Please change your trained .pkl file path and corresponding encoder type.

python3 adv_attack.py

Then you should add these new example to the original NLI dataset and keep training the NLI model for more epochs.

Citation

Please cite the following paper if the code or data help you:

@article{qin2020improving,
  title={Improving Sequence Modeling Ability of Recurrent Neural Networks via Sememes},
  author={Qin, Yujia and Qi, Fanchao and Ouyang, Sicong and Liu, Zhiyuan and Yang, Cheng and Wang, Yasheng and Liu, Qun and Sun, Maosong},
  journal={IEEE/ACM Transactions on Audio, Speech, and Language Processing},
  volume={28},
  pages={2364--2373},
  year={2020},
  publisher={IEEE}
}