Datasets:
Commit History
-
Update README.md
Verified signature7a462f4 -
Update README.md
Verified signaturec68febe -
Point metadata config at parquet (CSV builder was being applied to alignment.parquet)
Verified signature58c79bb -
Add metadata.parquet so both viewer configs share the parquet builder
Verified signaturea4b2117 -
Update README.md
Verified signature42f2a66 -
Fix alignment.parquet: performance_id as string, not Arrow dictionary (unblocks dataset viewer)
Verified signatureb5e56f5 -
Add dataset card: schema, statistics, limitations, ethics, licence status
Verified signature4c882e2 -
Upload 4 files
Verified signatureMAJEPPA is a corpus of 4,449 solo piano performance MIDI files (279 hours, ~5.0 million notes) from 3,979 recordings (303.2 hours), each labelled with the performer's expertise level and the recording context. 4,207 performances (95%) are paired with a reference score supplied as MIDI. The dataset was assembled to support research on piano performance. Its distinguishing feature is coverage of the full skill spectrum rather than the professional-recording bias typical of existing piano performance corpora. The release contains symbolic MIDI only. metadata.csv schema Column - performance_id: P_NNNN, matches a file in performance/ - recording_id: R_NNNN, groups performances cut from the same source video - score_id: S_NNNN, matches a file in score/. Empty when no score was matched - composer: Canonical composer name. Empty for 137 rows (traditional/popular works, or unattributed sources) - piece_title: Work title, including movement where applicable - recording_type: Performance context…faa681c
-
initial commit
Verified signature06c66fb