Speaker: Ming Jiang, PhD student at the Illinois Informatics Institute, University of Illinois at Urbana-Champaign
Time: Thursday, Nov 19 at 3 p.m. CST
Zoom: https://uwmadison.zoom.us/j/98435362731?pwd=T0xVem1mT1RnbWZzWmNaQ0s5dU1iUT09
The rapid growth of scientific articles has made digital libraries (DL) a vital source for scholarly information access. A grand challenge in DL research today is to organize large volumes of scholarly information to help researchers comprehend knowledge in their research fields. Academic knowledge graph, where scientific entities and relations are automatically extracted and linked, is a solution for this challenge.
In this talk, Ming Jiang will begin with an overview of relation mining from scholarly articles. Further, the talk will introduce a hybrid approach for extracting scientific relations, a critical step of constructing academic knowledge graphs, focusing on the semantic relationships between scientific terms. The method uses syntactic rules as a form of distant supervision to link related scientific term pairs to train BERT-based classifiers to identify relation pairs. The talk will include some examples of organizing scholarly knowledge into graphs at the document and corpus level. It will conclude with a discussion of some potential issues regarding the task set for future improvement.
Speaker: Ming Jiang is a PhD student in the Illinois Informatics Institute at the University of Illinois at Urbana-Champaign, working with Prof. J. Stephen Downie on research projects supported by the HathiTrust Research Center (HTRC). Her current research lies in Natural Language Processing (NLP) and Digital Libraries, aiming to make NLP techniques better comprehend information from digital objects such as eBooks and images.