Tamil

The 'Tamil' dataset/benchmark contains speech data used to evaluate automatic speech recognition (ASR) systems through fine-grained, Part-of-Speech (PoS)-wise error analysis, particularly focusing on the alignment of ASR hypotheses and reference transcriptions in non-Latin scripts.

Papers using Tamil (6)

MultiGen: Child-Friendly Multilingual Speech Generator with LLMs2025 · 3 citesBreaking the Script Barrier: Enabling Automatic Alignment for PoS-based ASR Error Analysis in Non-Latin Scripts2026Goodness-of-pronunciation without phoneme time alignment2026Dynamic Multi-Expert Projectors with Stabilized Routing for Multilingual Speech Recognition2026Textless NLP -- Zero Resource Challenge with Low Resource Compute2024Multistage Fine-tuning Strategies for Automatic Speech Recognition in Low-resource Languages2024