中央研究院 資訊科學研究所

Automatic knowledge base building is the key technology of constructing knowledge graph. It is a hot research topic in the Natural Language Processing (NLP) field. However, since the traditional named entities (NE) recognition only covers a few categories (such as Person, Location, Organization, etc.), it cannot satisfy those NLP applications which involve more categories. On the other hand, the type of open-domain entity is more flexible, and it thus can handle that kind of NLP tasks. The hypernyms (i.e., categories) of entity can be discovered from multiple sources. We learn the piecewise semantic projections from NEs to their hypernyms based on word embeddings, and then recognize the hypernym–hyponym relations automatically. With the constructed knowledge base, we further explore the document expanding task.

Bing Qin is a professor in Harbin Institute of Technology (HIT), and also a doctoral supervisor and the associate director of the Research Centre of Social Computing and Information Retrieval in HIT. Professor Bing Qin is the associate director of the field of Language and Knowledge Computing under Chinese Information Processing Society of China. Her research interests include natural language processing, text mining, information extraction, and sentiment analysis. She has published more than 60 papers in top conferences and journals, including ACL, COLING, EMNLP, IEEE TKDE and IEEE TASLP. She has been the area chair of various conferences. Also, she has been the project leaders for various Natural Science Foundation of China (NSFC) and National 863 projects. Currently, she is the project leader for the Chinese literature domain of the 863 project "Robots for College Entrance Examination" (高考機器人). She has won the first prize of WeiChang Qian Chinese Information Processing Scientific and Technological Award, and the second prize of Technological Invention of HeiLongJiang Province Award.