| The goal for this project is to explore strategies in adapting a Part of Speech (POS) tagger that was trained on Modern Standard Arabic sentences for tagging Levantine sentences, a dialect of Modern Standard Arabic, leveraged by methods of morphological analysis. I propose a tagging model that supports an explicit representation of the root-template patterns of Arabic. I will analyze the functionality and performance of the algorithms, and will compare the results. In leveraging the MSA POS tagger for tagging Levantine data, I achieved a peak accuracy of 73.28% which is 6% higher than the baseline for a standard Hidden Markov Model based tagger. |
