Page 72 - 2017 Brochure
P. 72
研究員
馬偉雲 Wei-Yun Ma
Assistant Research Fellow
Ph.D., Computer Science, Columbia University
Tel: +886-2-2788-3799 ext. 1819 Fax: +886-2-2782-4814
Email: ma@iis.sinica.edu.tw
http://http://www.iis.sinica.edu.tw/pages/ma/
• Assistant Research Fellow, Institute of Information Science, Academia Sinica (2014-present)
• Ph.D., Computer Science, Columbia University (2008-2014)
• M.S., Computer Science, Columbia University (2006-2008)
• Research Intern, Microsoft Research, United States (2013)
• Research Assistant, NLP group, Columbia University (2006-2014)
• Research Assistant, Institute of Linguistics, Academia Sinica (2006)
• Research Assistant, Institute of Information Science, Academia Sinica (2000-2006)
Research Description open domain knowledge bases, such as Freebase and Wikibase.
By using both context and knowledge bases for learned word
My research interests focus on machine learning in natural embeddings, our goal is to enhance existing NLP models and also
language processing (NLP), with an emphasis on the development make the embedding clearly interpretable.
of formal and computational models from a statistical perspective,
and the integration of linguistic knowledge in model design. Knowledge Acquisition from the Web
I also investigate how to acquire knowledge from the web. As
Distributional Word Representation it is known that the Internet contains an enormous amount of
One of my recent projects focuses on distributional word knowledge, my goal is to automatically acquire common sense or
representation, which is also called word embedding and is widely domain knowledge from the web by using our semantic parser as
used in NLP tasks. Compared with conventional symbolic word a tool to mine knowledge from text-based content. The knowledge
meaning representations, distributional word representation is is represented in the form of relations between objects, such as
trained from a corpus and represents meaning as vectors. This the Agent-Predicate Relation of human and eat, or Team-Member
representation provides improved computational capacity and the Relation of Nets and Jeremy Lin.
added advantage of generation. However, representations obtained
by unsupervised learning are short on explanation ability and do
not utilize prior knowledge bases that were previously constructed.
Therefore, we aim to improve the conventional context-based word
embedding process by incorporating prior knowledge bases from
semantic resources, such as WordNet or E-HowNet, and also from
Publications 7. Wei-Yun Ma and Kathleen McKeown, “Using a Supertagged
Dependency Model to Select a Good Translation in System
1. Hsin-Yang Wang, Wei-Yun Ma, “Integrating Semantic Knowledge Combination,” Proceedings of NAACL-HLT, July 2013.
into Lexical Embeddings Based on Information Content
Measurement,” Proceedings of Conference on European Chapter of 8. Wei-Yun Ma and Kathleen McKeown, “Detecting and Correcting
the Association for Computational Linguistics (EACL), April 2017. Syntactic Errors in Machine Translation Using Feature-Based
Lexicalized Tree Adjoining Grammars,” Proceedings of Conference
2. Jheng-Long Wu, Wei-Yun Ma, “A Deep Learning Framework for on Computational Linguistics and Speech Processing, October 2012.
Coreference Resolution Based on Convolutional Neural Network,”
International Conference on Semantic Computing, January 2017. 9. Wei-Yun Ma and Kathleen McKeown, “Phrase-level System
Combination for Machine Translation Based on Target-to-Target
3. Yu-Ming Hsieh, Wei-Yun Ma, “N-best Parse Rescoring Based on Decoding,” Proceedings of the 10th Biennial Conference of the
Dependency-Based Word Embeddings,” International Journal of Association for Machine Translation in the Americas, September
Computational Linguistics and Chinese Language Processing, volume 2012.
21, number 2, pages 19-34, December 2016.
10. Wei-Yun Ma and Kathleen McKeown, “Whereʼs the Verb Correcting
4. Hsin-Yang Wang, Wei-Yun Ma, “CKIP Valence-Arousal Predictor Machine Translation During Question Answering,” Proceedings of
for IALP 2016 Shared Task,” International Conference on Asian ACL-IJCNLP, July 2009.
Language Processing, November 2016.
5. Su-Chu Lin, Wei-Yun Ma and Yueh-Ying Shih, “A Unified
Representation of Attributes and Their Semantic Composition,”
In Proceedings of International Conference on Asian Language
Processing, November 2016.
6. Wei-Yun Ma and Kathleen McKeown, “System Combination
for Machine Translation through Paraphrasing,” Proceedings of
Conference on Empirical Methods in Natural Language Processing
(EMNLP), September 2015.
70 研究人員 Research Faculty
馬偉雲 Wei-Yun Ma
Assistant Research Fellow
Ph.D., Computer Science, Columbia University
Tel: +886-2-2788-3799 ext. 1819 Fax: +886-2-2782-4814
Email: ma@iis.sinica.edu.tw
http://http://www.iis.sinica.edu.tw/pages/ma/
• Assistant Research Fellow, Institute of Information Science, Academia Sinica (2014-present)
• Ph.D., Computer Science, Columbia University (2008-2014)
• M.S., Computer Science, Columbia University (2006-2008)
• Research Intern, Microsoft Research, United States (2013)
• Research Assistant, NLP group, Columbia University (2006-2014)
• Research Assistant, Institute of Linguistics, Academia Sinica (2006)
• Research Assistant, Institute of Information Science, Academia Sinica (2000-2006)
Research Description open domain knowledge bases, such as Freebase and Wikibase.
By using both context and knowledge bases for learned word
My research interests focus on machine learning in natural embeddings, our goal is to enhance existing NLP models and also
language processing (NLP), with an emphasis on the development make the embedding clearly interpretable.
of formal and computational models from a statistical perspective,
and the integration of linguistic knowledge in model design. Knowledge Acquisition from the Web
I also investigate how to acquire knowledge from the web. As
Distributional Word Representation it is known that the Internet contains an enormous amount of
One of my recent projects focuses on distributional word knowledge, my goal is to automatically acquire common sense or
representation, which is also called word embedding and is widely domain knowledge from the web by using our semantic parser as
used in NLP tasks. Compared with conventional symbolic word a tool to mine knowledge from text-based content. The knowledge
meaning representations, distributional word representation is is represented in the form of relations between objects, such as
trained from a corpus and represents meaning as vectors. This the Agent-Predicate Relation of human and eat, or Team-Member
representation provides improved computational capacity and the Relation of Nets and Jeremy Lin.
added advantage of generation. However, representations obtained
by unsupervised learning are short on explanation ability and do
not utilize prior knowledge bases that were previously constructed.
Therefore, we aim to improve the conventional context-based word
embedding process by incorporating prior knowledge bases from
semantic resources, such as WordNet or E-HowNet, and also from
Publications 7. Wei-Yun Ma and Kathleen McKeown, “Using a Supertagged
Dependency Model to Select a Good Translation in System
1. Hsin-Yang Wang, Wei-Yun Ma, “Integrating Semantic Knowledge Combination,” Proceedings of NAACL-HLT, July 2013.
into Lexical Embeddings Based on Information Content
Measurement,” Proceedings of Conference on European Chapter of 8. Wei-Yun Ma and Kathleen McKeown, “Detecting and Correcting
the Association for Computational Linguistics (EACL), April 2017. Syntactic Errors in Machine Translation Using Feature-Based
Lexicalized Tree Adjoining Grammars,” Proceedings of Conference
2. Jheng-Long Wu, Wei-Yun Ma, “A Deep Learning Framework for on Computational Linguistics and Speech Processing, October 2012.
Coreference Resolution Based on Convolutional Neural Network,”
International Conference on Semantic Computing, January 2017. 9. Wei-Yun Ma and Kathleen McKeown, “Phrase-level System
Combination for Machine Translation Based on Target-to-Target
3. Yu-Ming Hsieh, Wei-Yun Ma, “N-best Parse Rescoring Based on Decoding,” Proceedings of the 10th Biennial Conference of the
Dependency-Based Word Embeddings,” International Journal of Association for Machine Translation in the Americas, September
Computational Linguistics and Chinese Language Processing, volume 2012.
21, number 2, pages 19-34, December 2016.
10. Wei-Yun Ma and Kathleen McKeown, “Whereʼs the Verb Correcting
4. Hsin-Yang Wang, Wei-Yun Ma, “CKIP Valence-Arousal Predictor Machine Translation During Question Answering,” Proceedings of
for IALP 2016 Shared Task,” International Conference on Asian ACL-IJCNLP, July 2009.
Language Processing, November 2016.
5. Su-Chu Lin, Wei-Yun Ma and Yueh-Ying Shih, “A Unified
Representation of Attributes and Their Semantic Composition,”
In Proceedings of International Conference on Asian Language
Processing, November 2016.
6. Wei-Yun Ma and Kathleen McKeown, “System Combination
for Machine Translation through Paraphrasing,” Proceedings of
Conference on Empirical Methods in Natural Language Processing
(EMNLP), September 2015.
70 研究人員 Research Faculty
