Embedding
Short answer
An embedding is a list of numbers representing a piece of text's meaning, so that two passages saying the same thing in different words end up close together.
Embeddings are what lets a retrieval system match "termination" against a clause that says "cancellation". The text is converted to coordinates and the system looks for neighbours.
Nightjar does not currently use embeddings for retrieval — it matches words. That is a genuine limitation and it is listed as one: the upside is that it needs no second model and no index, the downside is that synonyms do not match.
Related terms
Other ways people ask this
- what is embedding