SDGCN
Modeling Sentiment Dependencies with Graph Convolutional Networks for Aspect-level Sentiment Classification
Install / Use
/learn @Pinlong-Zhao/SDGCNREADME
SDGCN
Tensorflow Implementations.
使用Tensorflow实现。
Requirement
- python 3.6 / 3.7
- tensorflow >= 1.12
- numpy
- sklearn
- bert_embedding
- See Bert Embeddings for more detail.
Usage
Creat the embedding
-
GloVe: Download pre-trained word vectors here. In this implement, we use glove.42B.300d.zip
-
BERT: Refer to creat_BERT_embedding.py and creat_BERT_embedding_2_targets.py to create BERT Embedding if need.
Training
python run_glove.py
Train model with GloVe Embedding. See run_glove.py for more training arguments.
python run_bert.py
Train model with BERT Embedding. See run_bert.py for more training arguments.
Citation
The manuscript is avaliable in arXiv:
"Modeling Sentiment Dependencies with Graph Convolutional Networks for Aspect-level Sentiment Classification". arXiv preprint arXiv:1906.04501 (2019) [pdf]
