String Ensemble Chords Dataset (SECD)

The String Ensemble Chords Dataset (SECD) is a large-scale collection of computationally generated string ensemble chord mixtures designed for music information retrieval (MIR), audio analysis, and machine learning research. It contains nearly 300,000 WAV files representing harmonic intervals, triads, and seventh chords across duo, trio, and quartet-like string configurations involving cello, viola, violin1, and violin2. SECD was constructed from isolated solo string recordings derived from the Philharmonia Orchestra Sound Samples library. The original source recordings are not redistributed; only the generated mixtures and associated metadata are included in this release. The released audio consists of generated multi-instrument mixtures, together with CSV metadata describing instrument assignments, pitches, dynamics, playing techniques, chord labels, voicing information, and source-reference information. The dataset is organized into separate downloadable archives covering harmonic intervals, triads, and seventh chords under loose and strict metadata-consistency conditions. SECD is intended to support tasks such as chord recognition, ensemble-size classification, dynamics classification, playing-technique classification, polyphonic audio classification, audio tagging, and related studies of controlled string-instrument mixtures.