説明
Knowledge Acquisition Team at RIKEN AIP
Speaker 1: Yuji Matsumoto
Title: Knowledge Acquisition from Scientific Papers
Abstract: Since the publication of research papers is increasing very rapidly, there is a strong demand for automatic analysis and extraction of information from scientific papers. This talk covers basic techniques of scientific paper analysis and introduces our activities of CREST and NEDO projects on information extraction from scientific papers.
Speaker 2: Noriki Nishida
Title: Analyzing discourse structure of scientific papers for knowledge acquisition
Abstract: Discourse structure represents how natural language text is organized locally to globally, and it has been proven to be useful in a variety of applications. However, discourse parsing still remains a challenge in NLP. In this talk, I will introduce our recent work on discourse parsing for scientific papers. I show that proxy-labeling algorithms can leverage unlabeled papers effectively for enhancing the quality of a discourse dependency parser, adapting it to a different domain, and reducing human annotation costs significantly (> 50%). I will also introduce our recent attempts to apply discourse parsing to knowledge acquisition from scientific papers.
Speaker 3: Hiroki Teranishi
Title: Neural Network Approaches to Coordination Disambiguation
Abstract: Coordination is a syntactic phenomenon in which two or more elements, known as conjuncts, are linked together typically by a coordinating conjunction. Coordinate structures frequently occur in natural language and are the major source of ambiguities. In this seminar, I introduce the importance of the research of coordination disambiguation and propose neural network-based methods to identify the conjuncts of coordinate structures.
