Newest papers

SyRA-CS: Synthetic-to-Real Adaptation for Code-Switched ASR with Meta-Learning

Fadel Thior, Ahmer Tabassum, Haania Siddiqui, Hanan Aldarmaki, Thamar Solorio · 2026 · arXiv

DOI: 10.48550/arXiv.2610.31415 · open-access PDF

Indonesian

Pengenalan Suara Otomatis (ASR) dengan pengalihan kode (CS) masih sulit karena suara CS nyata jarang dan suara sintetis tidak sepenuhnya cocok. Kami menyelidiki apakah suara CS sintetis lebih berguna untuk mempelajari inisialisasi yang ramah adaptasi daripada pengaturan halus konstan biasa. Kami mengusulkan SyRA-CS, kerangka kerja adaptasi sintetis-ke-realitas khusus pasangan berdasarkan meta-pelatihan Reptile. Untuk setiap pasangan bahasa dan model pra-pelatihan, kami melatih meta-latihan pada subset sintetis CS-FLEURS yang sesuai dan kemudian beradaptasi menggunakan satu jam suara target nyata yang ditandai. Kami mengevaluasi empat arsitek

Machine translation (Meta M2M100); the English original follows.

Abstract

Code-switched (CS) automatic speech recognition (ASR) remains difficult because real CS speech is scarce and synthetic speech does not fully match it. We investigate whether synthetic CS speech is more useful for learning an adaptation-friendly initialization than for ordinary continual fine-tuning. We propose SyRA-CS, a pair-specific synthetic-to-real adaptation framework based on Reptile meta-learning. For each language pair and pretrained model, we meta-train on the corresponding synthetic CS-FLEURS subset and then adapt using one hour of labeled real target speech. We evaluate four archite

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange