Published December 7, 2023
| Version v1
Dataset
Open
Spoken Gigaword
Description
Description
This is the synthetic Spoken Gigaword dataset, which are parts of the dataset created for the studies on interpreter-aided spoken language understanding (SLU) in the paper below, with three different parts:
- SLURP-Fr, an end-to-end SLU dataset based on the French portion of MASSIVE (https://github.com/alexa/massive), containing 16,521 synthetic audio samples created using Google TTS, accompanied with 477 real test samples collected from two French speakers at Idiap.
- SLURP -Es, a similar dataset based on the parallel Spanish portion of MASSIVE, containing only synthetic samples.
- Spoken Gigaword, a speech summarization dataset generated from Gigaword (https://www.tensorflow.org/datasets/catalog/gigaword), containing 51,385 synthetic audio samples created using Google TTS.
Reference
If you use this dataset, please cite the following publication:
He, Mutian, and Philip N. Garner. "The Interpreter Understands Your Meaning: End-to-end Spoken Language Understanding Aided by Speech Translation." Findings of EMNLP 2023.
Files
Files
(8.6 GB)
Name | Size | Download all |
---|---|---|
md5:eeae901af3ca6983125df32b8decac06
|
4.5 MB | Download |
md5:690b4acfa43cf7f42d13bd39d9802d79
|
8.6 GB | Download |
Additional details
Related works
- Is described by
- Conference paper: 10.48550/arXiv.2305.09652 (DOI)
Funding
- Storytelling and first impressions in face-to-face and algorithm-powered digital interviews 197479
- Swiss National Science Foundation