Dataset
Text-only English dataset of mixed dialog and non-dialog datasets, consisting of the following portions from the following works:
Spontaneous speech (4.2M tokens)
Staged dialogs (4.2M tokens)
Non-dialog (4.2M tokens)
Total 12.6M tokens (approximately 10M words)