本文已参与「新人创作礼」活动，一起开启掘金创作之路。

Neural Sentiment Classification with User and Product Attention

---

0.Information

Year : 2016 Authors : Huimin Chen, Maosong Sun, Cunchao Tu, Yankai Lin, Zhiyuan Liu Conference : EMNLP 2016

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.

---

1.Summary

This paper propose a model named NSC-UPA(Neural Sentiment Classification-User Product Attention) in document-level sentiment classification. This model Introduce user and product information based attension over different semantic levels(word level and sentence level) of a ducument. However, it do not consider user information and product information seperately.

---

2.Research Objective(s)

Propose a hierarchical neural network to incorporate global user and product information into sentiment classification.

---

3.Background / Problem Statement

• Most existing method only focus on the text content and ignore the crucial characteristics of users and products.
• (Tang et al., 2015b) bring in a text preference matrix and a representation vector for each user and product into CNN sentiment classifier. It modifies the word meaning in the input layer with the preference matrix and concatenates the user/product representation vectors with generated document representation before softmax layer. But there are two problems:
  • The introduction of preference matrix for each user/product is insufficient and difficult to be well trained with limited reviews.
  • The characteristics of user and product should be reflected on the semantic level besides the word level.

---

4.Method(s)

Model

NSC-UPA (Neural Sentiment Classification-User Product Attention)

---

Basic

• LSTM
• Attention mechanism
• Learning Semantic Representations of Users and Products for Document Level Sentiment Classification¹

---

5.Evaluation

Dataset

IMDB and Y elp Dataset Challenge in 2013 and 2014

---

Compared with baseline

Measure:

• Accuracy
• RMSE

---

Visualization

---

6.Conclusion

1. Propose an effective Neural Sentiment Classification model by taking global user and product information into consideration.
2. Introduce user and product information based attension over different semantic levels(word level and sentence level) of a ducument.

---

7.Notes

1. This paper do not consider user information and product information seperately.
2. Most users and products usually have some text information such as user and product profiles, they can be taken advantages of in sentiment analysis in future.
3. The effectiveness of this model on aspect level sentiment classification can be explored.

---

8.Code

github.com/thunlp/NSC

Footnotes

1. Tang, Duyu, Bing Qin, and Ting Liu. "Learning semantic representations of users and products for document level sentiment classification." Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2015. ↩