Token-level labels marking, in a voice-assistant query, the search term — the
minimal topic string you would hand to a knowledge base or search engine. Given
"what is the speed of light?" the target is "speed of light"; given
"set volume to fifty" the target is nothing (there is no topic to look up).
The task is not document keyphrase extraction and not full intent/slot
NLU. It answers one question: what do I search for? — the input the OVOS… See the full description on the dataset page:
https://huggingface.co/datasets/TigreGotico/search-term-extraction.
