A Generative Model of Phonotactics

A Generative Model of Phonotactics Richard Futrell Brain and Cognitive Sciences Massachusetts Institute of Technology futrell@mit.edu Adam Albright Department of Linguistics Massachusetts Institute of Technology albright@mit.edu Peter Graff Intel Corporation graffmail@gmail.com Timothy J. O’Donnell Department of Linguistics McGill University timothy.odonnell@mcgill.ca Abstract We present a probabilistic model of phono- tactics, the set of well-formed phoneme se- quences in a language. Unlike most compu- tational models of phonotactics (Hayes and Wilson, 2008; Goldsmith and Riggle, 2012), we take a fully generative approach, model- ing a process where forms are built up out of subparts by phonologically-informed struc- ture building operations. We learn an inven- tory of subparts by applying stochastic memo- ization (Johnson et al., 2007; Goodman et al., 2008) to a generative process for phonemes structured as an and-or graph, based on con- cepts of feature hierarchy from generative phonology (Clements, 1985; Dresher, 2009). Subparts are combined in a way that allows tier-based feature interactions. We evaluate our models’ ability to capture phonotactic dis- tributions in the lexicons of 14 languages drawn from the WOLEX corpus (Graff, 2012). Our full model robustly assigns higher proba- bilities to held-out forms than a sophisticated N-gram model for all languages. We also present novel analyses that probe model be- havior in more detail. 1 Introduction People have systematic intuitions about which se- quences of sounds would constitute likely or un- likely words in their language: Although blick is not an English word, it sounds like it could be, while bnick does not (Chomsky and Halle, 1965). Such in- tuitions reveal that speakers are aware of the restric- tions on sound sequences which can make up possi- ble morphemes in their language—the phonotactics of the language. Phonotactic restrictions mean that each language uses only a subset of the logically, or even articulatorily, possible strings of phonemes. Admissible phoneme combinations, on the other hand, typically recur in multiple morphemes, lead- ing to redundancy. It is widely accepted that phonotactic judgments may be gradient: the nonsense word blick is better as a hypothetical English word than bwick, which is better than bnick (Hayes and Wilson, 2008; Al- bright, 2009; Daland et al., 2011). To account for such graded judgements, there have been a vari- ety of probabilistic (or, more generally, weighted) models proposed to handle phonotactic learning and generalization over the last two decades (see Da- land et al. (2011) and below for review). How- ever, inspired by optimality-theoretic approaches to phonology, the most linguistically informed and suc- cessful such models have been constraint-based— formulating the problem of phonotactic generaliza- tion in terms of restrictions that penalize illicit com- binations of sounds (e.g., ruling out ∗bn-). In this paper, by contrast, we adopt a generative approach to modeling phonotactic structure. Our approach harkens back to early work on the sound structure of lexical items which made use of mor- pheme structure rules or conditions (Halle, 1959; Stanley, 1967; Booij, 2011; Rasin and Katzir, 2014). Such approaches explicitly attempted to model the 73 Transactions of the Association for …

Content truncated to 500 words