Meetup: Doing Hard Things with Less Data
Philip Blair, a Senior Research Engineer on our R&D team will be speaking at the Berlin Machine Learning Group. This meetup will be held online. Register here to learn about “Doing Hard Things with Less Data.”
Doing Hard Things with Less Data
Certain tasks in NLP lie at the unfortunate intersection between things which are hard (therefore requiring a significant amount of data) and things which are highly domain-specific (therefore limiting the ability to apply techniques such as transfer learning). In this talk, we explore one such task, event extraction, and discuss how instance-based techniques and active learning can be combined to combat these issues in a way that allows such complex models to be developed with surprisingly minimal data, thereby reducing the amount of labor and cost required to get a functioning model off the ground.
(40 min + 10 min Q&A)
Speaker: Philip Blair
Philip is a Senior Research Engineer at Basis Technology, based in Cambridge, Massachusetts, USA. He has spent the last 4 years applying cutting-edge NLP research to a large variety of industry needs, and he has published work on predicting downstream performance of word embeddings based on their intrinsic properties.