SLUE Tasks

Tasks

Task

Primary metric

Automatic Speech Recognition (ASR)

WER

Named Entity Recognition (NER)

F1

Sentiment Analysis

F1

Datasets

SLUE uses the VoxCeleb and VoxPopuli datasets.

We’ve diligently curated subsets of these datasets for fine-tuning and evaluation for SLUE tasks. You can take advantage of our redistribution, so you don’t need to download the entire (and large) dataset. With this dataset, we include the human annotation and transcription for SLUE tasks. All that’s required is to run the script and it will handle everything necessary - downloading and preprocessing included.

Here is a brief overview of the datasets. For more in-depth information, please refer to our paper.

Corpus Size - utts (hours) Tasks License
Fine-tune Dev Test
SLUE-VoxPopuli 5,000 (14.5) 1,753 (5.0) 1,842 (4.9) ASR, NER CC0 (check complete license here)
SLUE-VoxCeleb 5,777 (12.8) 1,454 (3.2) 3,553 (7.8) ASR, SA CC-BY 4.0 (check complete license here)