metadata
license: apache-2.0
raw-corpus is a reduced version from the tweets dataset
the dataset was made by extracting posts with the following criteria:
- at least 8 words
- at least 100 likes
- at least 70% of latin alphabet characters
just like the original dataset, this data has no filters, so expect your models to be as raw as they can