Identifying Primary Stress Across Related Languages and Dialects with Transformer-based Speech Encoder Models
This work addresses stress identification for speech processing in under-resourced languages, though it is incremental as it applies an existing transformer method to new data.
The paper tackled primary stress identification across related languages and dialects by fine-tuning a pre-trained transformer model with an audio frame classification head, achieving near-perfect results for Croatian and Serbian and a 10-point performance drop for more distant Chakavian and Slovenian.
Automating primary stress identification has been an active research field due to the role of stress in encoding meaning and aiding speech comprehension. Previous studies relied mainly on traditional acoustic features and English datasets. In this paper, we investigate the approach of fine-tuning a pre-trained transformer model with an audio frame classification head. Our experiments use a new Croatian training dataset, with test sets in Croatian, Serbian, the Chakavian dialect, and Slovenian. By comparing an SVM classifier using traditional acoustic features with the fine-tuned speech transformer, we demonstrate the transformer's superiority across the board, achieving near-perfect results for Croatian and Serbian, with a 10-point performance drop for the more distant Chakavian and Slovenian. Finally, we show that only a few hundred multi-syllabic training words suffice for strong performance. We release our datasets and model under permissive licenses.