Morphological Acquisition by Formal Analogy

While most approaches to unsupervised morphology acquisition often rely on metrics based on information theory for identifying morphemes, we describe a novel approach relying on the notion of formal analogy, that is, a relation between four forms, such as: reader is to doer as reading is to doing. Our assumption is that formal analogies identify pairs of morphologically related words, for instance reader/reading and doer/doing. Based on this assumption, our approach simply consists in identifying all the formal analogies involving the words in a lexicon. It turned out that for large lexicons, this happens to be a very time consuming task. Therefore, we report our attempts in designing practical systems based on the analogical principle. We applied our systems on the five languages of the shared task. The learning is made in an unsupervised manner, making use of the supplied lexicons only.