mastercaojie / pban-pytorch

A Position-aware Bidirectional Attention Network for Aspect-level Sentiment Analysis, PyTorch implementation.

PBAN-PyTorch

PyTorch implementation of Gu et al.'s COLING 2018 work.

Requirement

• PyTorch >= 0.4.0
• NumPy >= 1.13.3
• Python 3.6
• GloVe pre-trained word vectors:
  • Download pre-trained word vectors here.
  • Extract the glove.twitter.27B.zip and glove.42B.300d.zip to the \glove\ folder.

Dataset

Based on the restaurant and laptop dataset of SemEval-2014 Task 4.

Restaurant Dataset

Polarity	#Positive	#Negative	#Neutral
Train	2164	807	637
Test	728	196	196

Laptop Dataset

Polarity	#Positive	#Negative	#Neutral
Train	994	870	464
Test	341	128	169

Usage

Train the model:

python train.py --model_name pban --dataset restaurant

Show help message and exit:

python train.py -h

Implemented models

LSTM

Tang, Duyu, et al. "Effective LSTMs for Target-Dependent Sentiment Classification." Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers. 2016. [pdf]

ATAE-LSTM

Wang, Yequan, Minlie Huang, and Li Zhao. "Attention-based LSTM for aspect-level sentiment classification." Proceedings of the 2016 conference on empirical methods in natural language processing. 2016. [pdf]

PBAN

Gu, Shuqin, et al. "A Position-aware Bidirectional Attention Network for Aspect-level Sentiment Analysis." Proceedings of the 27th International Conference on Computational Linguistics. 2018. [pdf]

Performance

Restaurant Dataset

Three-class

Model	In Paper	This Code
LSTM	74.28	77.68
ATAE-LSTM	77.20	78.30
PBAN	81.16	80.89

Two-class

Model	In Paper	This Code
LSTM	-	-
ATAE-LSTM	90.90	90.26
PBAN	91.67	92.32

Laptop Dataset

Three-class

Model	In Paper	This Code
LSTM	66.45	71.00
ATAE-LSTM	68.70	71.32
PBAN	74.12	74.76

Two-class

Model	In Paper	This Code
LSTM	-	-
ATAE-LSTM	87.60	87.63
PBAN	87.81	87.42

Acknowledgements

• Some of the code is borrowed from songyouwei.
• Using this code means you have read and accepted the copyrights set by the dataset providers.

License

MIT