Objective per-mark feature columns derived from the companion silver corpus,
one row per application (serial_no, joins to
illeatmyhat/uspto-trademarks). Each column is a measurement of the
filing, not a judgement about any filer.
coined_mark_score
character n-gram improbability of the whole mark, [0,1] (1 = most coined/random-looking)
coined_token_max
max n-gram improbability over the mark's tokens, [0,1]… See the full description on the dataset page:
https://huggingface.co/datasets/illeatmyhat/uspto-trademarks-features.
