Misbah Al Mamun

Datasets

Data I built

Sylheti and Indian classical flute have almost no machine-readable resources. Each of these corpora was collected and curated from scratch because the research could not start without it.

21,181

aligned entries

Bangla–Syloti Nagri Parallel Corpus

Bangla sentences paired with Syloti translations in Bangla script and in Syloti Nagri script, for low-resource machine translation.

7,051

labelled sentences

SylEmotions

Dual-script Sylheti sentences labelled by native speakers across seven emotion classes: angry, happy, sad, hate, fear, surprise, neutral.

748 + 1,000

recordings + synthetic melodies

SwaraSet

Isolated swara recordings across the seven primary classes from professional flute musicians, over multiple octaves and playing styles, plus synthetic continuous melodies generated by DTW-based concatenation with time-stretching, carrying ground-truth boundary timestamps.