Adversarial Multi-Criteria Learning for Chinese Word Segmentation

Presentation on theme: "Adversarial Multi-Criteria Learning for Chinese Word Segmentation"— Presentation transcript:

1 Adversarial Multi-Criteria Learning for Chinese Word Segmentation
Xinchi Chen (Fudan University) Advisors: Prof. Xuanjing Huang Prof. Xipeng Qiu Direction: Natural Language Processing

2 What is Chinese word segmentation (CWS) ?
中国官员应邀到美国开会 中国/官员/应邀/到/美国/开会 大学生活好，还是中学生活好 大学/生活/好，还是/中学/生活/好 如果你饿了，我就下面给你吃 如果/你/饿/了，我/就/下/面/给/你/吃

3 基于词的方法 基于字的序列标注方法1 基于词和基于字的方法的结合 基于统计的分词
1. N. Xue Chinese word segmentation as character tagging. Computational Linguistics and Chinese Language Processing 8(1):29–48.

4 大学/生活/好/，/还是/中学/生活/好
基于字的序列标注方法 B E B E S S B E B E B E S 大学/生活/好/，/还是/中学/生活/好

5 Long Short-term Neural Network based CWS [X Chen, X Qiu, C Zhu, P Liu, X Huang; EMNLP 2015]

6 Adversarial Multi-Criteria Learning for Chinese Word Segmentation [X Chen, Zhan Shi, X Qiu, X Huang; ACL 2017]

7 Adversarial Multi-Criteria Learning for Chinese Word Segmentation [X Chen, Zhan Shi, X Qiu, X Huang; ACL 2017]

8 Adversarial Multi-Criteria Learning for Chinese Word Segmentation [X Chen, Zhan Shi, X Qiu, X Huang; ACL 2017]

9 Objective function

10 Unsupervised Domain Adaptation by Backpropagation [Yaroslav Ganin, et al.]

11 Adversarial Multi-Criteria Learning for Chinese Word Segmentation [X Chen, Zhan Shi, X Qiu, X Huang; ACL 2017]

12 Adversarial loss function
The criterion discriminator maximizes the cross-entropy of predicted criterion distribution p( |X) and true criterion. An adversarial loss aims to produce shared features, such that a criterion discriminator cannot reliably predict the criterion by using these shared features. Therefore, we maximize the entropy of predicted criterion distribution when training shared parameters. Unlike (Ganin et al., 2016), we use entropy term instead of negative cross-entropy.

13 Training

14 Experiments

15 Experiments

16 Experiments

17 Experiments

18 Experiments

19 Error Analysis

20 Case Study

21 Simplified Chinese to Traditional Chinese
Knowledge Transfer Simplified Chinese to Traditional Chinese Formal Texts to Informal Texts

22 Simplified Chinese to Traditional Chinese

23 Formal Texts to Informal Texts

24 Thank you for your attention!
Xinchi Chen (Fudan University) Advisors: Prof. Xuanjing Huang Prof. Xipeng Qiu Direction: Natural Language Processing