/KGAN

Primary LanguagePython

KGAN: Knowledge Graph Augmented Network Towards Multi-view Representation Learning for Aspect-based Sentiment Analysis

News

26/4/2022

  • Knowledge graph embeddings is added. Additionally, we update the implementation of KGNN-BERT and ASGCN-BERT. The corresponding training code is also updated.

Requirements

  • python 3.7
  • pytorch-gpu 1.7
  • numpy 1.19.4
  • pytorch_pretrained_bert 0.6.2
  • nltk 3.3
  • spacy 2.2.0
  • en_core_web_sm 2.1.0
  • GloVe.840B.300d
  • bert-base-uncased

Environment

  • OS: Windows 10
  • GPU: NVIDIA GeForce GTX 1660 SUPER
  • CUDA: 11.0
  • cuDNN: v8.0.4

Dataset

  • raw data: "./dataset/"
  • processing data: "./dataset_npy/"
  • word embedding file: "./embeddings/"
  • Knowledge graph embedding file: download from Google Drive

Training options

  • ds_name: the name of target dataset, ['14semeval_laptop', '14semeval_rest'], default='14semeval_rest'
  • bs: batch size to use during training, [32, 64], default=64
  • learning_rate: learning rate to use, [0.001, 0.0005], default=0.001
  • n_epoch: number of epoch to use, default=20
  • model: the name of model, default='KGNN'
  • dim_w: the dimension of word embeddings, default=300
  • dim_k: the dimension of graph embeddings, [200,400], default=200
  • is_test: train or test the model, [0, 1], default=1
  • is_bert: GloVe-based or BERT-based, [0, 1], default=0

Running

training based on GloVe:

  • python ./main.py -ds_name 14semeval_laptop -bs 32 -learning_rate 0.001 -n_epoch 20 -model KGNN -dim_w 300 -dim_k 400 -is_test 0 -is_bert 0
  • python ./main.py -ds_name 14semeval_rest -bs 64 -learning_rate 0.001 -n_epoch 20 -model KGNN -dim_w 300 -dim_k 200 -is_test 0 -is_bert 0

Evaluation

To have a quick look, we saved the best model weight trained on the evaluated datasets in the "./model_weight/best_model_weight". You can easily load them and test the performance. You can evaluate the model weight with:

  • python ./main.py -ds_name 14semeval_laptop -bs 32 -model KGNN -dim_w 300 -dim_k 400 -is_test 1
  • python ./main.py -ds_name 14semeval_rest -bs 64 -model KGNN -dim_w 300 -dim_k 200 -is_test 1

Notes

  • The datasets and more than 50% of the code are borrowed from TNet-ATT (Tang et.al, ACL2019).

Citation

@article{zhong2022knowledge,
  title={Knowledge Graph Augmented Network Towards Multiview Representation Learning for Aspect-based Sentiment Analysis},
  author={Zhong, Qihuang and Ding, Liang and Liu, Juhua and Du, Bo and Jin, Hua and Tao, Dacheng},
  journal={arXiv preprint arXiv:2201.04831},
  year={2022}
}