A token classifier gives you per-token tags. Nobody wants per-token tags: they want entities.
Given tokens and tags in BIO format, return the entities as a list of (text, label, start, end) tuples:
B-X begins an entity of type XI-X continues the entity of type X immediately before itO is outside any entitytext is the tokens joined by a single spacestart and end are token indices, end exclusiveReal taggers emit malformed sequences constantly, and how you handle them changes your reported F1. Two rules to apply:
I-X with no preceding entity starts a new entity of type XI-X following an entity of a different type starts a new entity rather than continuing the wrong oneBuild the architecture on a canvas: place the components, configure them, connect them into a data flow, and write a short reason for each one. The AI reviewer grades your design against a rubric written specifically for this problem.
Minimum 5 components · needs a wide desktop screen