Page 49 - profile2012.indd

Distinguished Chair and Distinguished Research Fellows 特聘講座/特聘研究員

許聞廉
特聘研究員

Wen-Lian Hsu
Tel: +886-2-2788-3799 ext. 1804
Fax: +886-2-2782-4814
Distinguished Research Fellow Email: hsu@iis.sinica.edu.tw
Ph.D., Operations Research, Cornell University http://iasl.iis.sinica.edu.tw/hsu/

Research Description Publications
研究介紹 代表著作

My main research interests include the following: graph Our main focus is on natural language understanding 1. W. L. Hsu and R. McConnell, “PC-trees and circular-ones arrangements”
algorithms, natural language understanding; and bioin- (NLU). We believe NER is most fundamental module in Theoretical Computer Science, 296(1), 99-116, (2003).
formatics. NLU since a good NE annotation would enable the simi- 2. W. F. Lu and W. L. Hsu. “A Test for the Consecutive Ones Property on Noisy
larity between two sentences to stand out. Along the Data - Application to Physical Mapping and Sequence Assembly,” Journal
In graph algorithms, we have done extensive and ground- way of building fundamental component modules, we of Computational Biology 10(5), 709-735, (2004).
breaking work on two fundamental classes of special have successfully developed a knowledge annotation 3. K. P. Wu, H. N. Lin, J. M. Chang, T. Y. Sung, and W. L. Hsu, “HYPROSP: ● B.S., mathematics, National Taiwan
graphs, namely planar graphs and interval graphs, and and inference kernel, InfoMap, for the semantic analysis A Hybrid Protein Secondary Structure Prediction Algorithm - a Knowledge- University, (1973)
Based Approach,” Nucleic Acids Research 32(17), 5059-5065, (2004).
their related graphs (such as circular arc graphs and circle of natural language, which can be applied to a wide va-
graphs). We introduced a new data structure [2], called riety of application systems in natural language process- 4. Hsin-Nan Lin, Jia-Ming Chang, Kuen-Pin Wu, Ting-Yi Sung and Wen-Lian ● M.S., Operations Research, Cornell
Hsu, “HYPROP II: A knowledge-based hybrid method for protein secondary
University, (1978)
PC-tree, which greatly simplifies the recognition of these ing, biological knowledge bases, automation of pipeline structure prediction based on local prediction confidence,” Bioinformatics

two classes of graphs. PC-tree algorithms can unify the experiments, and e-learning. Utilizing InfoMap, we won 21, 3227-3233, (2005). ● Ph.D., Operations Research, Cornell
tests of consecutive ones property and the circular ones the 1st place in the Chinese Question Answering contest 5. Kuen-Pin Wu, Jia-Ming Chang, Jun-Bo Chen, Chi-Fon Chang, Wen-Jin Wu, University, (1980)
property [10]. Furthermore, PC-tree is a natural represen- in NTCIR held in Tokyo, Japan in 2005 and 2007. Tai-Huang Huang, Ting-Yi Sung and Wen-Lian Hsu*, “RIBRA-an Error- ● Intelligent Chinese Input Software
tation for planar graph embedding. Our new planarity Tolerant Algorithm for the NMR Backbone Assignment Problem,” Journal -- Top ten Most Distinguished Chinese
of Computational Biology 13, 229-244, (2006).
test based on PC-trees is simple, elegant and yields a lin- Our natural language processing techniques have also computer products（十大傑出中文資
ear time algorithm for finding maximal planar subgraphs. been applied to biological literature mining. We have 6. Chuan-Yih Yu, Yin-Hao Tsui, Yi-Hwa Yian, Ting-Yi Sung, and Wen-Lian 訊產品獎）in Taiwan (1993)
Hsu*, “The Multi-Q Web Server for Multiplexed Protein Quantitation,” Nu-
successfully produced biological entity and relation rec- cleic Acids Research.35, W707-712, (2007).
In the field of natural language processing, we have de- ognition algorithms integrating biological knowledge ● Outstanding Research Awards（國
signed a Chinese input system called GOING, which auto- and machine learning techniques. In 2009, we won the 7. Allan Lo, Hua-Sheng Chiu, Ting-Yi Sung, Ping-Chiang Lyu, and Wen-Lian 科會傑出研究獎）by NSC (1991-92,
Hsu*, “Enhanced membrane protein topology prediction using a hierarchi-
94-95, 96-97)
matically translates a phonetic (or Pinyin) sequence into 1st place in BioCreative II.5 gene normalization task. In cal classification method and a new scoring function,” Journal of Proteome

characters with an accuracy rate close to 96%. It received addition, we have worked on gene-disease relation ex- Research 7, 487-496, (2008). ● K.T. Lee Research Breakthrough
the Distinguished Chinese Information Product Award（ traction for hypertension, obesity, and diabetes, which 8. Hong-Jie Dai, Chi-Hsin Huang, Ryan T. K. Lin, Richard Tzong-Han Tsai, Award（李國鼎穿石獎）(1999)
中文傑出資訊產品獎）in 1993. It is being used by more than will be extended to cover most diseases with a user- and Wen-Lian Hsu, “BIOSMILE web search: a web application for annotat- ● NSC Designated Research Fellow（國
ing biomedical entities and relations,” Nucleic Acids Research 36, W390-
a million people in Taiwan nowadays. feedback mechanism. W398, (2008). 科會特約研究員獎）(1999)
9. Chih-Chiang Tsou, Yin-Hao Tsui, Yi-Hwa Yian, Yi-Ju Chen, Chuan-Yih Yu, ● NSC Appointed Outstanding Re-
We have worked on several basic modules in Chinese nat- In bioinformatics, we have produced an error-tolerant Han-Yin Yang, Ke-Shiuan Lynn, Yu-Ju Chen, Ting-Yi Sung*, and Wen-Lian
ural language processing. Our research on Chinese word algorithm for the physical mapping and the clone as- Hsu*, “MaXIC-Q Web: A Fully Automated Web Service Using Statisti- search Award（國科會傑出特約研究
segmentation has won the 1st place in 2006 SIGHAN sembly problem. We have developed a suite of software cal and Computational Methods for Protein Quantitation Based on Stable 員獎）(2005)
CityU closed track competition. Identifying named en- for protein quantitation (Multi-Q, MaxiQ, and Ideal-Q), Isotope Labeling and LC-MS,” Nucleic Acids Research 37, W661-W669, ● Academia Sinica Investigator Award（
(2009).
tities, such as person, location, and organization names which has received a lot of attention and users. We have 中央研究院深耕獎）(2005)
in documents is very important for natural language un- also developed machine learning and knowledge-based 10. Chih-Chiang Tsou, Chia-Feng Tasi, Ying-Hao Tsui, Putty-Reddy Sudhir, Yi-
Ting Wang, Yu-Ju Chen, Jeou-Yuan Chen, Ting-Yi Sung, and Wen-Lian Hsu,
derstanding. In the past, we have developed a mature, algorithms for protein structure and function prediction “IDEAL-Q: An automated tool for label-free quantitation analysis using an ● IEEE Fellow (2006)

machine-learning based named entity recognition (NER) as well as protein interaction interface prediction. By efficient peptide alignment approach and spectral data validation,” Molecu- ● Teco Award（東元獎）(2008)
system, which won the second place in 2006 SIGHAN regarding protein sequences as a language and identi- lar & Cellular Proteomics 9, 131-144, (2010).
NER competition. fying protein synonyms from a sizeable database, we 11. Hsin-Nan Lin, Cédric Notredame, Jia-Ming Chang, Ting-Yi Sung*, Wen- ● Pan Wen Yuan Distinguished Research
developed a new sequence aligner, called SymAlign, to Lian Hsu*, “Improving the alignment quality of consistency based aligners Award (潘文淵研究傑出獎) ( 2010).
with an evaluation function using synonymous protein words,” PLoS ONE
improve sequence alignments. In testing several large 6(12): e27872, (2011). [IF 2010: 4.411].
dataset of protein pairs for structure similarity, it was 12. Mike Tian-Jian Jiang, Tsung-Hsien Lee, Wen-Lian Hsu, “The Left and Right
demonstrated that SymAlign has a much better preci- Context of a Word: Overlapping Chinese Syllable Word Segmentation with
sion than the other aligners (40% compared to 6% or less Minimal Context,” to appear in ACM Transaction on Asian Language Infor-
for the others). mation Processing.
特聘講座/特聘研究員
48 Distinguished Chair and Distinguished
Research Fellows
49