本文已参与「新人创作礼」活动，一起开启掘金创作之路。
Improving Review Representations with User Attention and Product Attention for Sentiment Classification
0.Information
Year : 2018 Authors : Zhen Wu, Xin-Yu Dai∗, Cunyan Yin, Shujian Huang, Jiajun Chen, Conference : AAAI-18
Wu, Zhen, et al. "Improving review representations with user attention and product attention for sentiment classification." Thirty-second AAAI conference on artificial intelligence. 2018.
1.Summary
This paper propose a new framework called HUAPA(Hierarchical User Attention and Product Attention neural network) which encode user information and product information seperately in document-level review sentiment classification. It performs much better on the Experiment-Dataset compared with baseline.
2.Research Objective(s)
Propose a novel framework to encode user and product infomation in document-level review sentiment classification.
3.Background / Problem Statement
- For different users, same word might express different emotional intensity.
- It is un- reasonable to encode user and product information together into one representation.
4.Method(s)
Model
HUAPA (Hierarchical User Attention and Product Attention neural network)
Basic
- Long Short-Term Memory
- Attention Mechanism
- Document-level Sentiment Classification
- NSC-UPA1
5.Evaluation
Dataset
IMDB and Y elp Dataset Challenge in 2013 and 2014
Compared with baseline
Measure:
- Accuracy
- RMSE
Visualization
6.Conclusion
- In this paper, we propose a novel framework incorporating user and product information for sentiment classification.
- the visualizations of attention also show our model can capture user information and product information.
7.Notes
- They encode user information and product information seperately.
- They did not use any regularization or dropout.
8.Code
Footnotes
-
Chen, Huimin, et al. "Neural sentiment classification with user and product attention." Proceedings of the 2016 conference on empirical methods in natural language processing. 2016. ↩
