Dish recognition
Dish recognition is the stage where GastroMatch understands what specific dish is present so that any further explanation or recommendation has a clear object.
Definition
Dish recognition is the process of identifying a dish from a photo or menu line strongly enough that GastroMatch can treat it as a concrete dish, not just an approximate label. The output is a named dish with a stable identity in the system.
This identity allows menu text, visual signals, and later user ratings to be attached to the same dish instead of drifting across vague categories.
Why it matters
- Connects perception to decision. Without knowing what dish is on the table or in the menu, the system cannot meaningfully explain or recommend.
- Enables consistent memory. Recognizing the same dish again lets GastroMatch use past feedback for future choices in the same or similar places.
- Supports local signals. Accurate recognition is the basis for building local dish signals around specific dishes in specific locations.
How GastroMatch uses it
- After the user takes a photo or highlights a menu item, GastroMatch first performs dish recognition before any interpretation or match scoring.
- The recognized dish becomes the anchor for Taste DNA reasoning, local dish statistics, and explanations shown to the user.
- When the user saves or rates the dish, that feedback is attached to the recognized dish entity, not just to a text string.
What it is not
- Not a full nutritional or cultural analysis of the dish — that belongs to interpretation.
- Not a generic “image search” or entertainment scan. The output is meant to be stable and decision-ready.
- Not a guarantee of confidence. If recognition is weak, the system should narrow, defer, or stay silent instead of guessing.
Internal links
To see where dish recognition appears in the broader product behavior: