Deciphering Assamese Vowel Harmony with Featural InfoWaveGAN

TL;DR

This paper explores how a novel Featural InfoWaveGAN model learns Assamese vowel harmony directly from raw speech data, capturing complex phonotactic patterns and feature preferences, advancing phonological understanding beyond text-based methods.

Contribution

The study introduces a new model that learns Assamese vowel harmony from raw speech, demonstrating its ability to grasp long-distance phonological patterns and feature preferences.

Findings

Model successfully learns iterative long-distance vowel harmony.

Produces speech error-like forms resembling language acquisition.

Identifies a specific vowel feature as a universal trigger.

Abstract

Traditional approaches for understanding phonological learning have predominantly relied on curated text data. Although insightful, such approaches limit the knowledge captured in textual representations of the spoken language. To overcome this limitation, we investigate the potential of the Featural InfoWaveGAN model to learn iterative long-distance vowel harmony using raw speech data. We focus on Assamese, a language known for its phonologically regressive and word-bound vowel harmony. We demonstrate that the model is adept at grasping the intricacies of Assamese phonotactics, particularly iterative long-distance harmony with regressive directionality. It also produced non-iterative illicit forms resembling speech errors during human language acquisition. Our statistical analysis reveals a preference for a specific [+high,+ATR] vowel as a trigger across novel items, indicative of feature learning. More data and control could improve model proficiency, contrasting the universality of learning.