Extracting the relations that exist between two entities is animportant step in numerous Web-related tasks such as information extraction. A supervised relation extraction system that is trained to extract a particular relation type might not accurately extract a new type of a relation for which it has not been trained. However, it is costly to create training data manually for every new relation typethat one might want to extract. We propose a method to adapt anexisting relation extraction system to extract new relation types withminimum supervision. Our proposed method comprises two stages:learning a lower-dimensional projection between different relations,and learning a relational classifier for the target relation type withinstance sampling. We evaluate the proposed method using a datasetthat contains 2000 instances for 20 different relation types. Ourexperimental results show that the proposed method achieves astatistically significant macro-average F-score of 62.77. Moreover,the proposed method outperforms numerous baselines and a previouslyproposed weakly-supervised relation extraction method.

Dr. Danushka Bollegala is an associate professor at the university ofTokyo. He obtained his BS, MS and PhD degrees from the university ofTokyo. Prior to joining the faculty at the university of Tokyo he wasa research fellow with the Japanese society for the promotion ofscience (JSPS). He has conducted research into several importantproblems in natural language processing and Web data mining such astext summarization, entity resolution (name disambiguation and aliasdetection), attributional and relational similarity measurement,relation extraction and domain adaptation of NLP systems. He haspublished his work in numerous international conferences such as WWW,ACL, IJCAI, ECAI and AAAI.