GSLT Machine Learning, Assignment 1: Machine Learning for Morphological Analysis
Atelach Alemu Argaw, Eva Forsbom · 2005
The purpose of this assignment is to define a learning task as described by Mitchell (1997), choose three algorithms from the book and describe how they could be usedapplied to the chosen learning problem. In many NLP applications, there is a need to look up words in a lexicon where only citation forms of each word are represented. Hence, morphological analysis (’lemmatization’) is a crucial step in such applications. Sometimes it is also important to be able to look up parts of a word, for example in the cases where the whole word is not represented in the lexicon (e.g. in MT). We considered the morphological analysis learning problem as a classification task, i.e. given the class, it should be possible to derive the citation form for a token, since the class contains all the information needed. The definition of the learning problem is as follows: