A community seed library wants each suggested variety to arrive with a one-sentence why this fits you blurb, not only a ranked ID list from a classical recommender. Which use case is this?
Select an answer to reveal the explanation.
Short Explanation
Think of a seed suggestion that comes with why this fits you, not just a ranked ID. That is a generative recommendation. Speech-to-text and a tax lookup do not write the blurb.
Full Explanation
Generative recommendations combine a suggestion with natural-language explanation—such as a one-sentence 'why this fits you' blurb—so the user understands the fit, not only a ranked identifier. The community seed library explicitly wants that explanatory copy with each suggested variety, which is the generative layer on top of (or instead of) a bare ranking list.
A ranked ID list only, with no sentence that explains the fit, is the classical recommender output pattern. It may order varieties well but fails the requirement for natural-language copy. Speech-to-text of the greenhouse radio transcribes audio into text; it does not recommend seeds or write fit blurbs. A statute-table lookup that computes an exact seed-packet tax is deterministic rule evaluation, not generative recommendation copy.
Exam caveat: on the exam, 'generative recommendation' is the tell when the stem asks for explanatory prose with the suggestion. Classical ranking alone is insufficient when the business asks for the why-it-fits sentence.
Operational check: sample recommended blurbs against actual grower constraints (sun, zone, season) and quarantine any suggestion whose prose invents traits not present in the seed catalog before publishing to patrons.