UUniversalLabeler
Loss-resistant annotation translation
v1.0.5 / Public spec

Language changes.
The event does not.

Pack the world, not just the sentence. UniversalLabeler turns an annotation into addressable claims—action, actors, hands, objects, state, place, time, outcome and evidence—then renders each language from the same packet.

01 / ObserveEN source
Synthetic first-person preview of two hands folding a pale blue towel on a dark table
FRAME 01:00:01.000—01:00:08.000
synthetic visual preview
Fold the towel on the table with both hands.
EVIDENCEhead_video · hand_pose
ENVIRONMENTindoor · laundry area
TIME SCOPE7.00 s · episode_clock
PROVENANCEsource-bound · reviewed
02 / PackUAP core v1
action object hands state Δ outcome evidence time place roles
CONTENT-BOUNDmeaning
packet
10 / 10 claims
10atomic claims
0silent drops
5projections
03 / Projectone packet → many surfaces
CLAIMS REPRESENTED10 / 10
CLAIMS OMITTED0
REVIEW STATECONFORMANCE
00 / Design principle

Compress the description.
Keep the world.

A universal vocabulary is not an infinite dictionary. It is a compositional address space. New actions and nouns become concept IDs; roles describe how they participate; evidence says why each claim is believed.

Fluent text is a view, not the database. English, Chinese, Tagalog, Spanish and Hindi can choose their own word order and grammar without becoming the authority for hands, state, timing or geometry.

Inside one event.

The visible sentence is short. The packet remains explicit. Click an event above and every value below updates from the same semantic frame.