[nl-uiuc] AIIS talk by Andreas Vlachos on April 8

[Rajhans Samdani]

Hi all!

This week we have Andreas Vlachos (http://www.biostat.wisc.edu/~vlachos/) from 
University of Wisconsin Madison. Details follow.

When: 2 pm, April 8 (Friday)

Where: 3405, Siebel Center

Title: Bridging the Gap between Pipelines and Structured Prediction in Biomedical 
Event Extraction

Abstract:
We develop a structured prediction approach to biomedical event extraction using the 
Searching to Learn framework which converts the problem of structured prediction into 
multiple cost-sensitive classification (CSC) tasks whose parameters are learnt jointly. 
We evaluate our performance on the BioNLP 2009 shared task datasets and show 
that SEARN improves the performance of a pipeline of classifiers learnt independently 
by 7 points in F-score, while maintaining its ease of implementation. In this process, 
we highlight some important aspects of the framework and in particular the choice of 
CSC learning algorithm.

Hoping to see y'all

Thanks!
Rajhans


Rajhans Samdani,
Graduate Student,
Dept. of Computer Science,
University of Illinois at Urbana-Champaign.