Skip to main content

Achievement

Unsupervised morphology learning

Research Achievements

Unsupervised morphology learning

IGERT Trainee Constantine Lignos, in collaboration with Professors Mitch Marcus (Computer Science) and Charles Yang (Linguistics) have been conducting extensive interdisciplinary work on computer-based language processing and language learning, inspired by human language learning and use. Their work in unsupervised morphology learning (i.e., learning stems, prefixes and suffixes) explores using simple, cognitively plausible statistics to learn a linguistically informed representation of the morphology of a language. They are also working on the problem of word segmentation in speech, exploring the power of cognitively-plausible approaches to the problem that focus on the learner's lexicon and structural, as opposed to distributional, cues in the input, such as prosody.

SEE MORE: