Danish Conversational and Read-aloud Speech Dataset (CoRal)

Innovation Fund Denmark has granted DKK 14 million to a project that will bring Danish speech technology to an international level. Over the next two years, we will develop a speech dataset called CoRal, which stands for Danish Conversational and read-aloud speech dataset.

The dataset will contain 1,000-1,500 hours of conversational and read-aloud speech from a broad and representative sample of the population in terms of gender, age, Danish dialects and foreign accents. At the same time, language models will be developed that can recognise Danish speech and read Danish text aloud.

All data and models will be tested and published on an ongoing basis so that developers, companies and public institutions can benefit from them from the start.