datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
parliament_hearings_processed
Preprocessed parliament hearings ASR dataset to truecased form.
Original dataset: https://lindat.mff.cuni.cz/repository/xmlui/handle/11234/1-3126
dataset_info:
features:
- name: id
dtype: string
- name: audio
dtype:
audio:
sampling_rate: 16000
- name: transcription
sequence: string
splits:
- name: train
num_bytes: 53645064353.18
num_examples: 191455
- name: test
num_bytes: 740331298.0
num_examples: 2726… See the full description on the dataset page: https://huggingface.co/datasets/jkot/parliament_hearings_processed.Orpheus_Hearing
Orpheus Dataset: Enhanced Audio-to-ABC Notation Conversion
This dataset was specifically designed to train models for converting audio signals into ABC music notation, leveraging a customized workflow and mutation mechanisms specially designed with music theory.
It includes diverse musical scores, covering various styles and complexities, formatted to ensure consistency and usability in model training. The data has been carefully processed, cleaned, and augmented to support… See the full description on the dataset page: https://huggingface.co/datasets/BOB12311/Orpheus_Hearing.HearInContextEnglish | 中文
HearInContext
A Benchmark for Implicit Context in Speech Recognition
Illustrative example: the same spoken request is disambiguated as flour or flower by different assistant histories. The dialogue and waveform are illustrative.
Same audio. Different contexts. Different meanings.
HearInContext is a Mandarin–English contextual speech recognition benchmark. It pairs the same audio with dialogue histories supporting different meanings to evaluate… See the full description on the dataset page: https://huggingface.co/datasets/OPPOer/HearInContext.HEAR-HSet
HEAR-HSet — HEAR Hierarchical Evaluation Set
面向零样本语音合成（Zero-shot TTS）的分层评测基准，覆盖基础泛化、副语言可控生成与高难复杂场景三大维度。
共包含 2,702 条 prompt-target 音频对，约 1GB，语言涵盖中文和英文。
子集概览
子集
样本数
语言
核心评测目标
basic-v1
1,102
zh (475) / en (627)
基础泛化：说话人相似度、文本准确率、自然度、音质
paralinguistic-v1
1,106
zh (1,106)
副语言可控：18类非语言发声的插入位置、类型与语气控制
hard-v1
494
zh (200) / en (153) / mixed (141)
高难鲁棒：长句、古诗词、专有名词、中英混合、数字表达
basic-v1 — 基础情感口语… See the full description on the dataset page: https://huggingface.co/datasets/dinosaaaur/HEAR-HSet.hear-benchmark-16khzHEAR-DS-16k
HEAR-DS Background Audio (16kHz)
Binaural background audio recordings from the HEAR-DS (Hearing Aid Research Database of Sounds) dataset, downsampled to 16kHz and chunked into 10-second segments for speech enhancement and acoustic scene classification research.
Dataset Description
This dataset contains background noise recordings from 7 acoustic environments, captured using in-the-canal (ITC) hearing aid microphones. Each sample includes stereo (left/right ear) audio.… See the full description on the dataset page: https://huggingface.co/datasets/nkdem/HEAR-DS-16k.circor-heart-soundhearing2translate-humeval
This repository contains the human evaluation experiment data for Hearing to Translate: The Effectiveness of Speech Modality Integration into LLMs 📄.
The code for the project is hosted at github.com/sarapapi/hearing2translate.
The annotations were collected using Pearmut (code), a lightweight platform that makes end-to-end human evaluation for multilingual tasks efficient and reliable.
The evaluations were done with bilingual speakers using the Pearmut tool with the MQM/ESA protocol… See the full description on the dataset page: https://huggingface.co/datasets/zouhar/hearing2translate-humeval.supreme-court-hearings-asr
Indian Supreme Court Hearings — ASR dataset
Sentence-level, force-aligned audio–text pairs from Indian Supreme Court hearings, prepared
for fine-tuning ASR models (e.g. Whisper). 46.9 hours across 23 hearings / 15 cases.
Load
from datasets import load_dataset
ds = load_dataset("kirandevraj/supreme-court-hearings-asr")
ds["test"][0] # {'audio': {'array', 'sampling_rate': 16000}, 'text': '...', ...}
Splits
split
clips
hours
train
24,422… See the full description on the dataset page: https://huggingface.co/datasets/kirandevraj/supreme-court-hearings-asr.Nat-HEAR-AmbisonicsHEAR
HEAR
🎉 EMNLP 2026 main conference 🎉
📄 arXiv ·
🌐 Project page ·
💻 Code ·
🤗 Dataset ·
🧠 Model
Hierarchical Evaluation of Attribution and Reasoning, a benchmark for
speaker-attributed understanding of multi-party speech.
Most speech benchmarks can be solved by transcribing the audio and reading the text. HEAR
cannot. Every question asks something about who is speaking, not only what is said,
and roughly half the benchmark comes… See the full description on the dataset page: https://huggingface.co/datasets/PleasedPenguin/HEAR.NSYNTH_PITCH_HEARheart-failure-audioHEARSpeakerCountIdentification_LibriCount-Fold2HeartAnomalyDetection_HeartbeatSoundsHEARSpeakerCountIdentification_LibriCount-Fold4HEAREnvironmentalSoundClassification_ESC50-Fold4HEARSpeakerCountIdentification_LibriCount-Fold3HEAREnvironmentalSoundClassification_ESC50-Fold1court-hearingHEAREmotionRecognition_CREMAD-Fold2HEAREnvironmentalSoundClassification_ESC50-Fold3heart-failure-deidentifiedHEARPercussionInstrumentsStrokeClassification_MridangamStroke-Fold1HEAREnvironmentalSoundClassification_ESC50-Fold2HEAREnvironmentalSoundClassification_ESC50-Fold5HEARVocalImitationClassification_VocalImitations-Fold2HEAREmotionRecognition_CREMAD-Fold3HEARPercussionInstrumentsStrokeClassification_MridangamStroke-Fold2HEARSoundEventRecognition_FSD50K