Jev vs BERT : classer, router, taguer avec le modèle de TypeSafe AI
Didier Sampaolo demonstrates Jev on support tickets, batch classification, and blog tags. He contrasts its zero-shot setup with a trained classifier and proposes testing alternative category trees before reorganizing a product catalog.
Original by LVLUPClassificationIntermediate17 min 51 secPublished
The demonstration supplies criteria for binary probabilities, category choices, and ordered scores without first labelling a training dataset.
Jev selects or scores supplied candidates rather than generating arbitrary text or discovering unrestricted entity names.
Didier proposes running classification simulations across competing product taxonomies before choosing a catalog structure.
Worth knowing
The presenter raises privacy and GDPR questions about sending customer text to a provider. This review does not establish the provider’s current hosting, contractual terms, or legal compliance.