danvics/soulsync
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 1

Refine noise removal in track title normalization

Browse source 
Updated the normalization logic to only remove metadata markers that do not indicate different recordings (e.g., 'radio edit', 'remaster'), while preserving markers like 'live', 'remix', 'acoustic', and 'instrumental' that represent distinct versions. This improves matching accuracy by ensuring only equivalent recordings are compared, delegating version differentiation to the matching engine.
This commit is contained in:
Broque Thomas 2026-01-09 14:17:36 -08:00
parent 5dab9e7259
commit 99c8f71420
1 changed files with 25 additions and 28 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

53
database/music_database.py
View file

@ -2070,56 +2070,53 @@ class MusicDatabase:
cleaned = re.sub(r'\s*[\]\)]\s*', ' ', cleaned) # Convert closing brackets/parens to space
cleaned = re.sub(r'\s*-\s*', ' ', cleaned) # Convert dashes to spaces too
# STEP 2: Remove version markers and noise patterns for better matching
# This allows "I Want You Back - Radio Edit" to match "I Want You Back"
# STEP 2: Remove metadata noise for better matching
# IMPORTANT: Only remove markers that describe the SAME recording with different metadata
# DO NOT remove markers that indicate DIFFERENT versions (live, remix, acoustic, etc.)
# Those are handled by the matching engine's version detection system
patterns_to_remove = [
# Basic markers
# Basic markers (content/parental ratings)
r'\s*explicit\s*', # Remove explicit markers
r'\s*clean\s*', # Remove clean markers
# Featuring/collaboration
# Featuring/collaboration (metadata, not different version)
r'\s*feat\..*', # Remove featuring
r'\s*featuring.*', # Remove featuring
r'\s*ft\..*', # Remove ft.
r'\s*with\s+.*', # Remove "with Artist"
# Edit versions
r'\s*radio\s+edit.*', # Remove "radio edit"
# Edit versions (same recording, different edit for format)
r'\s*radio\s+edit.*', # Remove "radio edit" - same song, radio format
r'\s*single\s+edit.*', # Remove "single edit" - same song, single format
r'\s*album\s+edit.*', # Remove "album edit" - same song, album format
r'\s*edit\s*$', # Remove trailing "edit"
r'\s*single\s+edit.*', # Remove "single edit"
r'\s*album\s+edit.*', # Remove "album edit"
# Live versions
r'\s*live\s+at.*', # Remove "live at..."
r'\s*live\s+from.*', # Remove "live from..."
r'\s*live\s+unplugged.*', # Remove "live unplugged"
r'\s*live\s*', # Remove standalone "live"
r'\s*unplugged.*', # Remove "unplugged"
# Remasters
# Remasters (same recording, different mastering)
r'\s*\d{4}\s*remaster.*', # Remove "2015 remaster"
r'\s*remaster.*', # Remove "remaster/remastered"
r'\s*remastered.*', # Remove "remastered"
# Version types
r'\s*original\s+version.*', # Remove "original version"
r'\s*album\s+version.*', # Remove "album version"
r'\s*single\s+version.*', # Remove "single version"
r'\s*extended\s+version.*', # Remove "extended version"
# Version clarifications (metadata, not different recordings)
r'\s*original\s+version.*', # Remove "original version" - clarification
r'\s*album\s+version.*', # Remove "album version" - clarification
r'\s*single\s+version.*', # Remove "single version" - clarification
r'\s*version\s*$', # Remove trailing "version"
# Soundtrack/source
# Soundtrack/source info (metadata about source)
r'\s*from\s+.*soundtrack.*', # Remove "from ... soundtrack"
r'\s*from\s+".*".*', # Remove "from 'Movie Title'"
r'\s*soundtrack.*', # Remove "soundtrack"
# Other common markers
r'\s*acoustic.*', # Remove "acoustic"
r'\s*instrumental.*', # Remove "instrumental"
r'\s*remix.*', # Remove "remix"
r'\s*demo.*', # Remove "demo"
]
# NOTE: We do NOT remove these - they indicate DIFFERENT recordings:
# - live, live at, live from, unplugged (different performance)
# - remix, mix (different mix)
# - acoustic (different arrangement)
# - instrumental (different version)
# - demo (different recording)
# - extended (different length/content)
# These are handled by matching_engine.similarity_score() which applies penalties
for pattern in patterns_to_remove:
cleaned = re.sub(pattern, '', cleaned, flags=re.IGNORECASE).strip()