Datasets:
metadata
tags:
- audio
- asr
- speech
- text
language:
- de
license: cc-by-4.0
de_asr_benchmark - German ASR Benchmark Dataset (test only)
This dataset aggregates multiple German speech test sets with paired transcripts. It is intended only as a benchmark and evaluation set, not for training or validation.
Split
- test: all examples combined from the listed sources.
Fields
- audio: audio file (various formats)
- text: transcript (plain text)
Sources
- De_Distant_Speech_Data_Corpus_1415_Realtek (test subset)
- Bundestag ASR - FT_test_dataset
- Common Voice 17 (German test set)