Google's Cloud Speech-to-Text transcriptions of the NST Dataset
This repository contains the testing part of the NST Acoustic database for Swedish (16 kHz).
Along with the original transcriptions, it also contains the predicted transcriptions as provided by Google's Cloud Speech-to-Text service. We hope that, by publishing this data to the public domain, other researchers can use this as a benchmark to evaluate their ASR systems against commercial services such as the one provided by Google.
The data can be found in the file nst-test.csv, which has the following columns:
- wav_filename: path to the sound file in the NST Testing set
- transcription: the original transcription provided in the NST Testing set
- google_transcription: the transcription predicted by Google's Speech-to-text service
The file paths in the wav_filename column assumes that the testing part of the NST dataset has been extracted into a directory named test. All paths are relative to the directory containing the test directory.