MINIMALLY SUPERVISED MACHINE LEARNING WORD SENSE DISAMBIGUATES TO AMHARIC TEXT

Abstract

The main objective of this thesis was to design word sense disambiguation (WSD) minimally final model for Amharic words semi- upervised learning is half way between the supervised and unsupervised learning. In addition to unlabeled data, the algorithm is provided with some supervision information but not necessarily for all example data. Due to the unavailability of Amharic word net. Only five words and have their fifteen classes or senses were selected. These words were mesal, ras, yemigeba, gb, kena, a different data sets using three meanings or sense of words were prepared for the development of this Amharic WSD prototype. The final classification work was done on fully labeled training set using RBF network and bayes net classification algorithms on weka package.