Treat 'type beat' as a wrong-version keyword

A "type beat" is an instrumental track produced in another artist's
style, uploaded to SoundCloud and tagged with that artist's name to
game search ranking. They show up as candidates for major-label
tracks (e.g. "Eminem - Greatest (Kamikaze) Type Beat - Sit Down" for
"Greatest" by Eminem) and have nothing to do with the real song.

Add 'type beat' to the version-keyword list so the scorer applies the
0.4x penalty + flags the result as wrong_version. Currently the
matcher rejects them via low text-similarity scores anyway, but the
explicit keyword makes the rejection deterministic and gives a clear
diagnostic in the logs / modal.
pull/495/head
Broque Thomas 2 weeks ago
parent 2aff3dc210
commit da424d4bf6

@ -89,7 +89,12 @@ def get_valid_candidates(results, spotify_track, query):
# Detect if the expected track is a specific version (live, remix, acoustic, etc.)
expected_title_lower = (expected_title or '').lower()
_version_keywords = ['remix', 'live', 'acoustic', 'instrumental', 'radio edit',
'extended', 'slowed', 'sped up', 'reverb', 'karaoke']
'extended', 'slowed', 'sped up', 'reverb', 'karaoke',
# Producer-tag noise common on SoundCloud — "type
# beat" is an instrumental track produced in
# someone's style, tagged with the artist name to
# game search. NEVER the real song.
'type beat']
expected_is_version = any(kw in expected_title_lower for kw in _version_keywords)
scored = []

Loading…
Cancel
Save