On August 16-17, Stony Brook University hosted the Mathematics of Language (MOL) conference, where current Rutgers graduate student Tatevik Yolyan gave a talk. Her talk was titled “A Framework for Learning Phonological Maps as Logical Transductions,” and the proceedings paper can be accessed at this link. The abstract is copied below.
Abstract:
This paper presents an approach to learning a subclass of phonological maps that can be expressed with non-recursive quantifier-free logical transductions within the framework of Boolean Monadic Recursive Schemes (BMRS). Building on previous work on phonological learning with partially-ordered hypothesis spaces, this paper shows that similar structures can be used in learning phonological maps when these maps are represented by logical transducers over unconventional string models. A primary contribution of this paper is a demonstration of how the BMRS framework can be used to learn phonological generalizations.