New: Cookbooks and AI ExplanationsStep-by-Step recipes to solve problems connected to Roadmaps and Cheat Sheets. Need more details? Use AI buttons for structured and simple explanations with concrete examples throughout the whole platform.Take a look
Our support logs have the same question typed five different ways, and exact-match dedup misses every one of them.
What you'll have at the end
A ranked list of your dataset's most similar sentence pairs, with the true duplicates removed by hand.
You need
A list of short text strings pulled from your own dataset (support tickets, survey answers, FAQ questions), plus a working call that turns one piece of text into one embedding vector.
Not covered
Merging duplicates automatically with no human read, or scaling this same review past a few thousand sentences with an approximate-nearest-neighbor index; this recipe scores and reviews a batch small enough to read pair by pair.
Leans on
Support tickets sorted into topics without hand labeling every one
go there instead if you're sorting a whole ticket backlog into named themes rather than collapsing near-identical repeats of one question
Choose cosine similarity or dot product for your vectors
go there for the fuller reasoning behind matching your distance metric to your vectors; this recipe assumes cosine similarity and applies it
Verify your first embedding works
go there first if you don't yet have a working call that turns one piece of text into one vector
Checked 15 Aug 2026
Part of the Embeddings cookbook