A compact development set for video retrieval, claim extraction, and report
generation. It uses the same schema as the larger multivent-raw, so scripts
that target one transfer straight to the other.
This dataset card covers the core release: videos, audio, keyframes, and
the public evaluation annotations. Derived signals (OCR text, ASR transcripts,
visual / audio / video / omni embeddings) live in a companion release,
microvent-features, with its own dataset card… See the full description on the dataset page:
https://huggingface.co/datasets/hltcoe/microvent.
