long form
griot-duplex-en-longform-v1
GRIOT Duplex v1
Two-channel (full-duplex) synthetic call-centre conversations, built for training and
evaluating full-duplex spoken dialogue models — models that listen and speak at the
same time rather than taking strict turns.
Each file is one complete conversation with one speaker per channel, so overlaps,
interruptions, backchannels ("mm-hm", "right") and silences are preserved exactly as they
occur in time. Single-channel corpora lose this.
The repository holds two distinct… See the full description on the dataset page: https://huggingface.co/datasets/moiesmud/griot-duplex-en-longform-v1.leaderboard_longformasr-leaderboard-longform
ASR Leaderboard: Longform Test Sets
This dataset provides three longform ASR benchmark test sets — Earnings-21, Earnings-22, and TED-LIUM — used in the Hugging Face ASR Leaderboard.
Each dataset is provided in a standardized, Parquet format, ensuring safe, efficient loading through the Hugging Face datasets library.
This collection is designed for evaluating longform automatic speech recognition (ASR) models under real-world conditions such as extended audio segments… See the full description on the dataset page: https://huggingface.co/datasets/hf-audio/asr-leaderboard-longform.LongForm
LongForm
The LongForm dataset is created by leveraging English corpus
examples with reverse instructions. We select a
diverse set of human-written
documents from existing corpora such as C4 and
Wikipedia and generate instructions for the given
documents via LLMs. Then, we extend these examples with structured corpora examples such as Stack Exchange and WikiHow and task examples such as question answering, email writing, grammar error correction, story/poem… See the full description on the dataset page: https://huggingface.co/datasets/akoksal/LongForm.tedlium-long-form
Dataset Card for "tedlium-long-form"
To create the dataset:
import os
import numpy as np
from datasets import load_dataset, DatasetDict, Dataset, Audio
import soundfile as sf
from tqdm import tqdm
tedlium = load_dataset("LIUM/tedlium", "release3")
merged_dataset = DatasetDict()
validation_speaker_ids = [
"Al_Gore",
"Barry_Schwartz",
"Blaise_Agueray_Arcas",
"Brian_Cox",
"Craig_Venter",
"David_Merrill",
"Elizabeth_Gilbert",
"Wade_Davis",
]… See the full description on the dataset page: https://huggingface.co/datasets/distil-whisper/tedlium-long-form.earnings22-longform-with-ids
Earnings 22 longform test dataset with an index
Exact same content as the dataset used in the open ASR leaderboard here (split earnings22), but with the additional column "id".
We used the index of each sample as an id.