Computational Perspectives on Phonological Constituency and Recursion

Whether or not phonology has recursion is often conflated with whether or not phonology has strings or trees as data structures. Taking a computational perspective from formal language theory and focusing on how phonological strings and trees are built, we disentangle these issues. We show that even considering the boundedness of words and utterances in physical realization and the lack of observable examples of potential recursive embedding of phonological constituents beyond a few layers, recursion is a natural consequence of expressing generalization in phonological grammars for strings and trees. While prosodically-conditioned phonological patterns can be represented using grammars for strings, e.g., with bracketed string representations, we show how grammars for trees provide a natural way to express these patterns and provide insight into the kinds of analyses that phonologists have proposed for them.

Normalish Amenable Subgroups of the R. Thompson Groups

Results in [Formula: see text] algebras, of Matte Bon and Le Boudec, and of Haagerup and Olesen, apply to the R. Thompson groups [Formula: see text]. These results together show that [Formula: see text] is non-amenable if and only if [Formula: see text] has a simple reduced [Formula: see text]-algebra. In further investigations into the structure of [Formula: see text]-algebras, Breuillard, Kalantar, Kennedy, and Ozawa introduce the notion of a normalish subgroup of a group [Formula: see text]. They show that if a group [Formula: see text] admits no non-trivial finite normal subgroups and no normalish amenable subgroups then it has a simple reduced [Formula: see text]-algebra. Our chief result concerns the R. Thompson groups [Formula: see text]; we show that there is an elementary amenable group [Formula: see text] [where here, [Formula: see text]] with [Formula: see text] normalish in [Formula: see text]. The proof given uses a natural partial action of the group [Formula: see text] on a regular language determined by a synchronising automaton in order to verify a certain stability condition: once again highlighting the existence of interesting intersections of the theory of [Formula: see text] with various forms of formal language theory.

Modelling a subregular bias in phonological learning with Recurrent Neural Networks

A number of experiments have demonstrated what seems to be a bias in human phonological learning for patterns that are simpler according to Formal Language Theory (Finley and Badecker 2008; Lai 2015; Avcu 2018). This paper demonstrates that a sequence-to-sequence neural network (Sutskever et al. 2014), which has no such restriction explicitly built into its architecture, can successfully capture this bias. These results suggest that a bias for patterns that are simpler according to Formal Language Theory may not need to be explicitly incorporated into models of phonological learning.

Getting down to basics

2020 ACM A.M. Turing Award recipients Alfred Aho and Jeffrey Ullman helped develop formal language theory, invented efficient algorithms to drive the tasks of a compiler, and put them all together in 'The Dragon Book.'

The Future of Computational Linguistics: On Beyond Alchemy

Over the decades, fashions in Computational Linguistics have changed again and again, with major shifts in motivations, methods and applications. When digital computers first appeared, linguistic analysis adopted the new methods of information theory, which accorded well with the ideas that dominated psychology and philosophy. Then came formal language theory and the idea of AI as applied logic, in sync with the development of cognitive science. That was followed by a revival of 1950s-style empiricism—AI as applied statistics—which in turn was followed by the age of deep nets. There are signs that the climate is changing again, and we offer some thoughts about paths forward, especially for younger researchers who will soon be the leaders.