Phonological Fossils: Machine Learning Detection of Non-Mainstream Vocabulary in Sulawesi Basic Lexicon

TL;DR

This study uses machine learning to identify non-mainstream vocabulary in Sulawesi languages, revealing phonological patterns and geographic distribution without evidence of a single substrate language.

Contribution

It combines rule-based cognate subtraction with a phonological classifier to detect non-conforming vocabulary, advancing computational methods in historical linguistics.

Findings

Identified 438 candidate substrate forms (26.5%) in Sulawesi languages.

The classifier achieved an AUC of 0.763 in distinguishing non-mainstream forms.

No evidence found for a single pre-Austronesian language layer.

Abstract

Basic vocabulary in many Sulawesi Austronesian languages includes forms resisting reconstruction to any proto-form with phonological patterns inconsistent with inherited roots, but whether this non-conforming vocabulary represents pre-Austronesian substrate or independent innovation has not been tested computationally. We combine rule-based cognate subtraction with a machine learning classifier trained on phonological features. Using 1,357 forms from six Sulawesi languages in the Austronesian Basic Vocabulary Database, we identify 438 candidate substrate forms (26.5%) through cognate subtraction and Proto-Austronesian cross-checking. An XGBoost classifier trained on 26 phonological features distinguishes inherited from non-mainstream forms with AUC=0.763, revealing a phonological fingerprint: longer forms, more consonant clusters, higher glottal stop rates, and fewer Austronesian prefixes. Cross-method consensus (Cohen's kappa=0.61) identifies 266 high-confidence non-mainstream candidates. However, clustering yields no coherent word families (silhouette=0.114; cross-linguistic cognate test p=0.569), providing no evidence for a single pre-Austronesian language layer. Application to 16 additional languages confirms geographic patterning: Sulawesi languages show higher predicted non-mainstream rates (mean P_sub=0.606) than Western Indonesian languages (0.393). This study demonstrates that phonological machine learning can complement traditional comparative methods in detecting non-mainstream lexical layers, while cautioning against interpreting phonological non-conformity as evidence for a shared substrate language.