Page 93 - profile2014.indd

研究員
蘇克毅 Keh-Yih Su

Research Fellow
Electrical Engineering, University of Washington
Tel: +886-2-2788-3799 ext. 1801 Fax: +886-2-2782-4814
Email: kysu@iis.sinica.edu.tw
http://www.iis.sinica.edu.tw/pages/kysu

Research Description ● Research Fellow, IIS, Academia
Sinica (2014/6 - )
My research interests include Machine Translation, Machine Reading, Chinese Language ● President/CEO, Behavior Design
Processing, Natural Language Understanding, Statistical Language Modeling and Machine Corporation (1998/1 - 2014/5)
Learning. I have worked on statistical modeling and machine learning for natural language ● Associated Professor, Professor,
processing since 1986 (mainly adopting principled approaches to solving NLP problems, in Dept. of Elect. Eng., National
contrast to the prevailing techniques of simply adding a lot of features under the Maximum Tsing Hua Univ. (1984/8 - 1998/1)
Entropy framework). ● Ph.D., Electrical Engineering,
Univ. of Washington, Seattle, USA
My past experience mainly lies in building a statistical semantic machine translation sys- (1984)
tem. I proposed and built a statistics-oriented framework for machine translation, which ● Advisor, “New Series in Natural
automatically obtains associated parameters from the given pairs of Source and Target Language Processing”, John Ben-
Semantic-Normal-Forms via semi-supervised learning. In addition, I proposed a joint model jamins (1999 -)
to integrate Translation Memory into a phrase-based machine translation system during ● President, Asian Federation of

decoding, which achieved signi cant improvement. Natural Language Processing
(2011 - 2012)
I have also worked on Chinese word segmentation with new generative models (also with
additional resources later), and have obtained the best performance in the literature on ● Executive Member, Association
SIGHAN close/open-test (which is classi ed based on whether additional resources (such as for Computational Linguistics

(2005 - 2007)
dictionaries) other than the training-set data could be adopted).
● President, Association for
I will start to build an Intelligent Q&A System which is able to give desired answers that Computational Linguistics and
cannot be directly retrieved from the given text database. In other words, we would like to Chinese Language Processing
(1996 - 1997)
tackle the problem that inference is required in obtaining the answer (or the result must be
aggregated from information distributed in various locations). In addition to the answer, the

associated inference chain (and the corresponding con dence level) will also be given. This
project is expected to start with reading elementary school texts, before gradually shifting to high school texts, and then to real domain-
oriented applications. The progress of this project will be evaluated by comprehensive tests associated with the given text.

Publications

1. Kun Wang, Chengqing Zong, and Keh-Yih Su, Aug. 25-29, 6. Xiaoqing Li, Kun Wang, Chengqing Zong, and Keh-Yih Su,
2014, “Dynamically Integrating Cross-Domain Translation Dec. 8-15, 2012, “Integrating Surface and Abstract Features
Memory into Phrase-Based Machine Translation during De- for Robust Cross-Domain Chinese Word Segmentation”, Pro-
coding”, Proceedings of COLING 2014, Dublin, Ireland. ceedings of COLING 2012, Mumbai, India, pages 1653-1669.
2. Xiaoqing Li, Chengqing Zong, and Keh-Yih Su, “A Unified 7. Kun Wang, Chengqing Zong, and Keh-Yih Su, “Integrating

Model for Solving the OOV Problems of Chinese Word Seg- Generative and Discriminative Character-Based Models for
mentation,” Accepted by ACM Transactions on Asian Lan- Chinese Word Segmentation,” ACM Transactions on Asian
guage Information Processing. Language Information Processing, Vol.11, No. 2, Article 7,
June 2012, pages 7:1-7:41.
3. Xiaoqing Li, Chengqing Zong, and Keh-Yih Su, Nov. 22-14,

2013, “A Study of the Effectiveness of Suffixes for Chinese 8. Yufeng Chen, Chengqing Zong, and Keh-Yih Su, “An Interac-
Word Segmentation”, Proceedings of PACLIC 2013, Taipei, tion Approach to Recognizing and Aligning Chinese-English
Taiwan. Bilingual Named Entities,” Chinese Journal of Computers,
Vol.34, No. 9, 2011, pages 1688-1696.
4. Kun Wang, Chengqing Zong, and Keh-Yih Su, Aug. 4-9, 2013,
“Integrating Translation Memory into Phrase-Based Machine 9. Kun Wang, Chengqing Zong, and Keh-Yih Su, Aug. 23-27,
Translation during Decoding”, Proceedings of ACL 2013, So- 2010, “A Character-Based Joint Model for Chinese Word Seg-

fia, Bulgaria. mentation”, Proceedings of COLING 2010, Beijing, China,
pages 1173-1181.
5. Yufeng Chen, Chengqing Zong, and Keh-Yih Su, “A Joint
Model to Identify and Align Bilingual Named Entities,” Jour- 10. Yufeng Chen, Chengqing Zong, and Keh-Yih Su, July 11–16,
nal of Computational Linguistics, Vol.39, No. 2, June 2013, 2010, “On Jointly Recognizing and Aligning Bilingual Named
pages 229-266. Entities”, Proceedings of ACL 2010, Uppsala, Sweden.

93