Datasets:
Over-filtering problem with current banned.txt
Pre-1900 futurists writings about the future century may be banned.
Why it matters? Common practice is to test time-capsule LLMs asking about well known 20 century phenomena (not exact facts/surnames/brands). Current year-filters inhibit such prognostic texts important for answering future-century related questions.
I don't see this as a huge problem. People talking about future are very likely to use the same words other people are using, they won't really invent new words that will trip the pre-1900 filter.
The year filter is disabled when I filter datasets such as this one, I keep the other ones more aggressive values to compensate.
This text is sure to be vintage because I only used sources from HuggingFace, there are no secret sources, so I only used the pre-1900 filter as a precaution.