Learning k-ISL functions with phonological features

Abstract

This study investigates the grammatical inference of phonological processes from featural representations, focusing on the class of k-Input Strictly Local (k-ISL) functions (Chandlee et al. 2014; Chandlee and Heinz 2018). We present the Factor-by-Feature (FxF) framework which decomposes the learning task into smaller ones along featural dimensions. The FxF framework is compatible with any grammatical inference algorithm designed to infer string-to-string functions in terms of finite-state transducers. The Structured Onward Subsequential Inference Algorithm (SOSFIA) is one such grammatical inference algorithm which has proven learnability results in linear time and data (Jardine et al. 2014). We empirically evaluate FxF on five phonological processes across four languages: English, Yawelmani, Chukchi, and Polish, representing cases of feature spreading, feature changing, and feature insertion. Comparisons are provided in each case between learning these processes over segments with SOSFIA alone and with SOSFIA in the FxF framework. To better measure data efficiency, we also introduce AMβA, a sampling algorithm that identifies smaller characteristic samples which, informally, are sets of data that guarantee successful generalization. Our results demonstrate that FxF consistently yields more compact and learnable data representations, significantly reducing the size of characteristic samples. These findings underscore the importance of phonological representation and support linguistically informed learners for phonological grammars.