Memory-based part-of-speech tagging of Tagalog

• المؤلفون الرئيسيون: Trogo-Oblena, Rhia S., Raga, Rodolfo C., Jr.
• التنسيق: text
• منشور في: Animo Repository 2006
• الموضوعات: Natural language processing (Computer science); Text processing (Computer science); Computational linguistics—Philippines; Computer Sciences; Physical Sciences and Mathematics; Software Engineering
• الوصول للمادة أونلاين: https://animorepository.dlsu.edu.ph/faculty_research/12872
• المؤسسة: De La Salle University

We explore the application of memory-based learning to tagging of Tagalog Text. Memory-based learning is a form of supervised classification-based learning method based on similarity-based reasoning. It entails building a set of cases in memory based on feature-value patterns extracted from a manually tagged corpus and extrapolating the part-of-speech tag of a given new word in a particular context from the most similar cases in memory. In this paper, we discuss the architecture of MBTPOST, a memory-based Tagalog POS tagger, and present the output of an experiment using MBTPOST on a small training corpus, particularly looking at the benefits of shifting window sizes as well as assigning positional weights value to words occurring in different window positions to address the fluidity of word structure in Tagalog sentences.