Dr. Prachya Boonkwan
National Electronics and Computer Technology Center, Thailand
Learning with Intuitions: Semi-Supervision in Natural Language Processing
Abstract: Machine learning plays an important role in mo
dern natural language processing. Producing such rich linguistic resources for supervised learning is labor- and time-intensive, requiring well-trained linguists to define and annotate such linguistic structures and resolve any inconsistent annotations. Unsupervised learning has gained general interest for several decades, as it offers a possibility of building a statistical NLP module from raw text and reduces the labor of constructing linguistic resources from scratch. In this keynote speech, I will focus on how to incorporate linguistic intuitions into unsupervised learning to circumvent the issue of unexpected frequent collocation. The results on multilingual syntactic parsing show that linguistic intuitions can improve the parsing accuracy in all languages.
Prachya Boonkwan received B.Eng. and M.Eng. degrees in Computer Engineering from Kasetsart University in 2002 and 2005, respectively. He received a Ph.D. degree in Informatics from the University of Edinburgh, UK, in 2014. Since 2005, he has been working as a research for Language and Semantic Technology Lab at NECTEC, Thailand. His topics of interest include grammar induction, statistical parsing, statistical machine translation, natural language processing, machine learning, and formal syntax.