Wu, Yunfang is an associate professor in the School of Computer Science. She obtained her B.Sc. from Southwest University in 1995, and Ph.D. from Peking University in 2003 respectively. Her research interests include natural language processing, language resource construction, discourse parsing and question answering.
Dr. Wu has published more than 50 research papers, and some of them are published in top-tier conferences and journals, such as COLING, LRE (Language Resources and Evaluation), JCST (Journal of Computer Science and Technology). She also published one book concerning text analysis. She has served as the reviewer of various conferences, including COLING, NLPCC, CCL, CLSW, and she is the PC Co-Chair of CLSW 2017. She got the Second Prize of State Science and Technology Awards in 2011 as the eighth author.
Dr. Wu has been responsible for more than eight research projects including NSFC, NSSFC, etc., and she also participated in two 973 programs and three 863 projects. Her research achievements are summarized as follows:
1) Chinese language resource construction. She proposed the Chinese corpus annotation schema for word sense disambiguation and discourse parsing, and constructed a large corpus annotated with word senses and a large corpus annotated with discourse relations. She also built various knowledge bases of Chinese lexical semantics. Based on these language resources, she organized various semantic evaluation tasks, including SemEval-2007 Task 5, SemEval-2010 Task 18, SemEval-2012 Task 4, NLPCC-2016 Task 3 and NLPCC-2017 Task 1.
2) Natural language processing methods. She made a deep analysis on Chinese coordinate structures, and proposed a rule-based method to automatically annotate Chinese coordinate structures. She proposed the principle for distinguishing Chinese word senses, and employed an ensemble method to do word sense disambiguation. She applied the distributional hypothesis to compute Chinese lexical similarities, and also employed the Newman algorithm to cluster similar words in a graph model. She proposed a ranking-like SVM (SVM-R) model for sentence-level Chinese discourse tree building, and also she proposed several methods to incorporate discourse relation knowledge to the task of sentiment analysis. For question answering, she applied a novel multi-level gated recurrent neural network (GRNN) with non-textual information to predict the dialog act tag, and she proposed an original tensor neural network to model the relevance between question and answer and proposed a novel denoising tensor autoencoder (DTAE).
