Learning Count Classifier Preferences of Malay Nouns
Jeremy Nicholson, Timothy J. Baldwin · 2008
We develop a data set of Malay lexemes labelled with count classifiers, that are attested in raw or lemmatised corpora. A maximum entropy classifier based on simple, languageinspecific features generated from context tokens achieves about 50 % F-score, or about 65 % precision when a suite of binary classifiers is built to aid multi-class prediction of headword nouns. Surprisingly, numeric features are not observed to aid classification. This system represents a useful step for semisupervised lexicography across a range of languages. 1