Monday, November 14
12.10 - 1.00
CSE 4140

Speaker: Mark Land, Parity Computing
Title: High-Accuracy Author Disambiguation in Scientific Literature

Abstract:
Author Disambiguation is the problem of determining which articles in a corpus were written by the same author. Disambiguation must deal with the fact that multiple authors may share the same name (e.g. John Smith) and, conversely, that a given author may publish under multiple names (e.g. Bill vs. William). Often there is limited data with which to make a determination (e.g. an author's initial rather than full first name). We describe an iterative, knowledge-based system utilizing multiple lines of evidence to construct author profiles on collections of published scientific articles. On a corpus of over 40M articles our system has achieved over 99% precision and 98% recall, which compares very favorably with other disambiguation systems.