WVU NLP Class Participation in ShARe/CLEF Challenge

V. Jagannathan,D. Ganesan,A. Jagannathan,R. Kavi,A. Lamb,F. Peters,S. Seeger

Abstract:The spring 2013 graduate class in NLP decided to participate in the ShARe/CLEF challenge Tasks 1 & 2. The timing for the challenge coincided nicely with the spring semester session. There were six students in the graduate class, and the challenge tasks appeared to be good material to expose students to a practical task faced by healthcare industry. The general approach used by the class is to use CRF learning algorithm using Factorie – a scala-language based toolkit. The F-scores for best results for Task 1a relaxed – 0.801, 1a strict – 0.554, 1b relaxed – 0.625 & 1b strict 0.349, respectively. Task 2 was attempted as a group task. F-scores were 0.426 and 0.428 for strict and relaxed respectively. It was a real challenge to focus on the challenge as a class project. The students did learn how to apply NER CRF engine to a practical problem.