TASTE

Terroir-Aligned Storytelling & Text Engine

LLM -Knowledge Graph – Food & Wine – Explainable AI

TASTE (Terroir-Aligned Storytelling & Text Engine) is a controlled LLM-assisted authoring pipeline for producing domain-grounded digital content for food & wine enterprises. The tool is designed around structured inputs (company–product–territory–experience profiles) and a template-driven generation layer that produces consistent, reusable content packages across channels (web pages, product sheets, social posts, short video scripts, QR micro-copy). TASTE does not aim at unconstrained “creative generation”: it converts verified enterprise knowledge (raw materials, production steps, timing constraints, sensory descriptors, territorial identity, visit formats) into channel-specific textual assets under explicit editorial rules.

The system enforces guardrails to reduce hallucinations and to preserve factual integrity: generation is constrained by the declared profile, forbidden/allowed claims, and hard rules (e.g., no undisclosed certifications, no invented quantitative statements, no absolute superlatives). Outputs are versioned and traceable to the input fields they were derived from, enabling review workflows and iterative refinement. From an engineering standpoint, TASTE supports configurable bundles (e.g., VISIT, PRODUCT, SOCIAL/VIDEO) and exposes machine-readable exports (e.g., YAML/JSON + rendered text assets) for integration with websites, content management systems, or downstream analytics/decision-support modules.

Contacts:
Prof. Luca Pulina