||Feb 24, 2016
||Dr. Ryo Yoshinaka, Assistant Professor, Graduate School of Informatics, Kyoto University
Distributional learning of some context-free languages and beyond
This talk is concerned with the algorithmic learning of context-free and mildly context-sensitive languages.
In these year approaches generically called "distributional learning” have made great success in the learning of CFLs.
Distributional learning algorithms do Not learn probability distribution.
Distributional learning is interested in which combination of a string and a context form a sentence belonging to the concerned CFLs, where a context is a pair of strings that wraps a string in the middle to form a complete sentence.
We will review some of the highlights of the development of distributional learning theory and generalization of distributional learning techniques to mildly context-sensitive languages, which are proper but mild extensions of CFLs.