Cross-Domain Adaptation of Spoken Language Identification for Related Languages: The Curious Case of Slavic Languages
This addresses domain adaptation for language identification in related languages, which is incremental as it builds on existing neural LID systems.
The study tackled the problem of domain mismatch in neural spoken language identification for six Slavic languages, finding that out-of-domain speech severely hinders performance and that spectral features are more robust than cepstral features under domain shift, with unsupervised domain adaptation achieving relative accuracy improvements of 9% to 77%.
State-of-the-art spoken language identification (LID) systems, which are based on end-to-end deep neural networks, have shown remarkable success not only in discriminating between distant languages but also between closely-related languages or even different spoken varieties of the same language. However, it is still unclear to what extent neural LID models generalize to speech samples with different acoustic conditions due to domain shift. In this paper, we present a set of experiments to investigate the impact of domain mismatch on the performance of neural LID systems for a subset of six Slavic languages across two domains (read speech and radio broadcast) and examine two low-level signal descriptors (spectral and cepstral features) for this task. Our experiments show that (1) out-of-domain speech samples severely hinder the performance of neural LID models, and (2) while both spectral and cepstral features show comparable performance within-domain, spectral features show more robustness under domain mismatch. Moreover, we apply unsupervised domain adaptation to minimize the discrepancy between the two domains in our study. We achieve relative accuracy improvements that range from 9% to 77% depending on the diversity of acoustic conditions in the source domain.