Danish Speech Recognition Audio and Text Evaluation Set

Comprising just above 5 hours of Danish speech audio (.wav files) with corresponding reference text, created by Alvenir to evaluate automatic speech recognition models. It includes recordings from 50 speakers aged 20-60 years.

Use Cases

• Evaluate ASR model performance on Danish speech using .wav files and reference text transcripts.
• Analyze speech recognition accuracy across a speaker cohort of 50 individuals with varying ages.
• Fine-tune or adapt existing ASR models for Danish using a limited set of audio-text pairs.

Strengths

• Contains just above 5 hours of audio data, providing a quantifiable evaluation corpus.
• Includes recordings from 50 distinct speakers, offering speaker diversity.
• Provides corresponding reference text for each audio file, enabling supervised evaluation.

Limitations

• The amount of data is very limited, with just over 5 hours, which is insufficient for training robust models.
• Specific details on audio quality, recording conditions, or text transcription accuracy are not provided in the input.

Provenance

Source

Alvenir.

Collection Method

Collected by a third-party vendor through their software and people.

Freshness

Last updated on 2025-05-16.

Geography

Denmark (implied by Danish language).