This work uses unsupervised learning to chart ASL morphemes and detect them in natural signing.
In the learning phase, we utilize SignASL and ASL-LEX to extract robust phonological representations, identify patterns, and find correlates with meaning.
In the application phase, we will transfer this method to whole sentences to the end of translating classifiers and invented signs.
Presently about halfway through the learning phase. More details to come.