Brigitte Mathiak

Junior Professor for Digital Humanities, University of Cologne

 

"SMAFIRA - Saving Animal from Suffering with Text Mining"

Abstract

In Germany, animal protection legislation requires scientists to use alternatives methods to animal experiments, should such methods be available to pursue their scientific goals. In practise, many scientists are willing to use alternative methods, but finding such methods is surprisingly difficult. The difficulty lies not so much in identifying which methods make animals suffer more or less, but in finding relevant, yet method-agnostic literature for a "scientific goal". Together with our research partners at the Federal Bureau for Risk Assessment, we ran some preliminary analysis, which shows that it is indeed very difficult to find "goal"-related literature based on keyword queries. However, with Machine Learning, based on weighted term vectors, we were able to increase Precision from 13% to an estimated 95% with both Naive Bayes and SVM approaches. We are now exploring ways to diminish the required number of training examples by using interactive feedback methods and are hoping to find domain-independant linguistical properties of "goals" that would allow us to transfer information learned in one use case to another.

 

Research Interest: Prof. Mathiak is newly-appointed junior professor for Digital Humanities. She currently works on a variety of projects centering around using Text Mining and Information Retrieval in interdisciplinary research. At the workshop, she will explain how better search engines can save (animal) lives. Her other interests are entity extraction and assisting text annotation with both human intelligence and automatisms.