Evaluating a Phonotactic Learner for MITSL-(2,2) Languages
- Jacob K. Johnson (University of Utah)
- Aniello De Santo (University of Utah)
Abstract
We provide an implementation of De Santo and Aksënova (2021) \'s grammatical inference learning algorithm for Multiple Input-sensitive Tier-based Strictly Local languages (De Santo and Graf, 2019) — following the standard of SigmaPie (Aksënova, 2020), and evaluate it on an array of patterns with varying degrees of (subregular) complexity. MISTL languages are able to capture the interaction of local and non-local constraints, and while also handling multiple dependencies simultaneously. Their practical learnability thus has strong implications for the viability of grammatical inference/subregular approaches to phonotactic learning broadly. Additionally, the transparency and provable correctness of the learning algorithms developed for such formal classes can be of help in probing properties of phonotactic corpora more generally.
Keywords: Formal Language Theory, Subregularity, Tiers, Learnability
How to Cite:
Johnson, J. K. & De Santo, A., (2023) “Evaluating a Phonotactic Learner for MITSL-(2,2) Languages”, Society for Computation in Linguistics 6(1), 379-382. doi: https://doi.org/10.7275/crgk-6g04
Downloads:
Download PDF