First dedicated ASR corpus of 66 hours and systematic benchmarks for Puno Quechua using participatory collection and open release of data and fine-tuned models.
hub
Omnilingual asr: Open-source multilingual speech recognition for 1600+ languages
18 Pith papers cite this work. Polarity classification is still indexing.
18
Pith papers citing it
hub tools
citation-role summary
method 1
citation-polarity summary
years
2026 18roles
method 1polarities
use method 1representative citing papers
UrduSpeech is a 156-hour high-fidelity Urdu speech corpus with 12-dimension paralinguistic annotations, a 9-hour manually corrected benchmark, and open-source release to support speech technology for an under-resourced language.
Phoneme-based interfaces match or surpass projector-based ones for LLM ASR, especially in low-resource languages, and a BPE-phoneme hybrid offers additional improvements.
Multilingual ASR models show 39.7-297% zero-shot WER on Pashto public data, Whisper models output correct script in under 0.8% of cases, and fine-tuned models degrade to 32.5-59% WER on out-of-domain sets.
Vaani Benchmark V1.0 is a multimodal Hindi ASR dataset from 104 districts featuring spontaneous speech recordings in real-world conditions and three independent transcriptions per segment for robust multi-reference evaluation.
UR-BERT scales multilingual TTS encoders to 495 languages via Romanization unification and speech token prediction, outperforming baselines with better generalization.
MADE is a new multilingual agentic diagnosing engine that produces higher-quality diagnostic reports (47% better than baseline) on a large-scale evaluation substrate covering 33 model families and 26 languages.
JSPG jointly combines semantic, pinyin, and glyph retrieval with an extended Smith-Waterman algorithm to dynamically filter keyword dictionaries and improve accuracy in Chinese contextual ASR.
AfriVox-v2 is a benchmark that evaluates modern speech models on in-the-wild African audio with domain-specific tests for sectors including government, finance, health, and agriculture.
BlasBench supplies an Irish-aware normalizer and scoring harness that enables reproducible ASR comparisons and exposes a 33-43 point generalization gap for fine-tuned models versus 7-10 points for massively multilingual ones.
OmniVoice introduces a diffusion language model-style non-autoregressive TTS system that directly maps text to multi-codebook acoustic tokens, scaling zero-shot synthesis to over 600 languages with SOTA results on multilingual benchmarks using 581k hours of open data.
Evaluation of WhisperIPA and ZIPA reveals persistent performance gaps across languages, accents, gender, ethnicity, and age even after allowing for similar phoneme substitutions.
SALSA adapts speech-aware LLMs via supervised layer-wise steering vectors, reporting up to 46.8% relative gains over zero-shot on out-of-domain speech benchmarks.
Introduces INSV-A automated screening benchmark for Pashto TTS systems reporting WER, script fidelity, and LID results across five systems on FLEURS and Common Voice prompts.
Case study finds that fine-tuned ASR models outperform human listeners on Dutch dysarthric continuous speech from one speaker, lowering WER from over 70% to over 23%.
Fine-tuned Wav2Vec2 and HuBERT models recognize click consonants more accurately than non-clicks in G|ui and West !Xoon data.
A narrative survey of low-resource NLP evaluation identifies the Annotation Scarcity Paradox as a structural mismatch between scalable models and scarce sociolinguistic evaluation capacity.
citing papers explorer
-
Building Community-Centred NLP Resources for Puno Quechua
First dedicated ASR corpus of 66 hours and systematic benchmarks for Puno Quechua using participatory collection and open release of data and fine-tuned models.
-
Phonemes vs. Projectors: An Investigation of Speech-Language Interfaces for LLM-based ASR
Phoneme-based interfaces match or surpass projector-based ones for LLM ASR, especially in low-resource languages, and a BPE-phoneme hybrid offers additional improvements.
-
Vaani Benchmark V1.0: An Inclusive Multimodal Benchmark Dataset for Hindi
Vaani Benchmark V1.0 is a multimodal Hindi ASR dataset from 104 districts featuring spontaneous speech recordings in real-world conditions and three independent transcriptions per segment for robust multi-reference evaluation.
-
UR-BERT: Scaling Text Encoders for Massively Multilingual TTS Through Universal Romanization and Speech Token Prediction
UR-BERT scales multilingual TTS encoders to 495 languages via Romanization unification and speech token prediction, outperforming baselines with better generalization.
-
MADE: Beyond Scoring via a Multilingual Agentic Diagnosing Engine for Fine-Grained Evaluation Insights
MADE is a new multilingual agentic diagnosing engine that produces higher-quality diagnostic reports (47% better than baseline) on a large-scale evaluation substrate covering 33 model families and 26 languages.
-
JSPG: Dynamic Dictionary Filtering via Joint Semantic-Pinyin-Glyph Retrieval for Chinese Contextual ASR
JSPG jointly combines semantic, pinyin, and glyph retrieval with an extended Smith-Waterman algorithm to dynamically filter keyword dictionaries and improve accuracy in Chinese contextual ASR.
-
AfriVox-v2: A Domain-Verticalized Benchmark for In-the-Wild African Speech Recognition
AfriVox-v2 is a benchmark that evaluates modern speech models on in-the-wild African audio with domain-specific tests for sectors including government, finance, health, and agriculture.
-
OmniVoice: Towards Omnilingual Zero-Shot Text-to-Speech with Diffusion Language Models
OmniVoice introduces a diffusion language model-style non-autoregressive TTS system that directly maps text to multi-codebook acoustic tokens, scaling zero-shot synthesis to over 600 languages with SOTA results on multilingual benchmarks using 581k hours of open data.
-
Evaluating Bias in Phoneme-Based Automatic Speech Recognition Systems: An Analysis of IPA Transcription Models
Evaluation of WhisperIPA and ZIPA reveals persistent performance gaps across languages, accents, gender, ethnicity, and age even after allowing for similar phoneme substitutions.
-
SALSA: Speech Aware LLM Adaptation via Learned Steering Activation Vectors
SALSA adapts speech-aware LLMs via supervised layer-wise steering vectors, reporting up to 46.8% relative gains over zero-shot on out-of-domain speech benchmarks.
-
PashtoTTS-Bench: automated screening for low-resource non-Latin-script text-to-speech
Introduces INSV-A automated screening benchmark for Pashto TTS systems reporting WER, script fidelity, and LID results across five systems on FLEURS and Common Voice prompts.
-
Comparing Human and Automatic Recognition of Dutch Dysarthric Continuous Speech: A Case Study
Case study finds that fine-tuned ASR models outperform human listeners on Dutch dysarthric continuous speech from one speaker, lowering WER from over 70% to over 23%.
-
Pretrained self-supervised speech models can recognize unseen consonants
Fine-tuned Wav2Vec2 and HuBERT models recognize click consonants more accurately than non-clicks in G|ui and West !Xoon data.
-
The Annotation Scarcity Paradox in Low-Resource NLP Evaluation: A Decade of Acceleration and Emerging Constraints
A narrative survey of low-resource NLP evaluation identifies the Annotation Scarcity Paradox as a structural mismatch between scalable models and scarce sociolinguistic evaluation capacity.