Anton on transducer continuity (09/13)

There has been lots of work over the last ~20 years situating the phonotactics of natural language in certain classes of subregular languages and finding logical, automata-theoretic and algebraic characterizations of these classes.

This has allowed linguists not only to better understand the cognitive nature of phonological generalizations but also to design learning algorithms for them, something that is provably intractable if the regular space is not restricted in some way.

More recently there has been an increased effort to use similar methodology in the study of phonological processes. Ideally, the results linguists have uncovered while studying phonotactics would be transferable into this new domain, and indeed this has been possible in many cases.

Continuous transductions allow us to lift certain classes of languages to classes of string-to-string functions in a way that has not been explored from a linguistic perspective yet. We will talk about what it means for a transduction to be continuous, how this notion relates to numerical continuous functions and more.