Pronoun Resolution in Optimality Theory: An Automated System for Dutch Pronouns
Emiel de Munck
The vastly increasing amount of available human generated data and the automated collection thereof, bring a growing need for automated yet accurate Natural Language Processing (NLP). The nature of many modern Natural Language Processing systems is probabilistic, often relying on (big) data and self-learning techniques. Although these methods are of high practical value, a more transparent solution may be preferred in order to gain true linguistic insight on the matter. In this research, we examine and rebuild a previous attempt (by G. Bouma) in Optimality Theory for the problem of solving pronouns in Dutch. We then proceed to look for opportunities of improvement by diving into recent developments and theories, as well as by extending the experiment. The conclusion is that we could indeed improve the original project, yet our solution does not outperform the more popular methods.
Emiel de Munck, "Pronoun Resolution in Optimality Theory: An Automated System for Dutch Pronouns", Master's Thesis for the Media Technology programme, Leiden University (The Netherlands), 2014