# Domain vocabulary the transcriber cannot know. # # One term per line. "#" starts a comment. A term may contain spaces and is matched as a phrase. # Terms shorter than 4 characters are ignored - too easy to collide with ordinary words. # # ONLY DISTINCTIVE TERMS BELONG HERE: system names, site names, operators, product codes. Do NOT add # ordinary Czech words. "balíček" is one edit from "malíček" (little finger), so listing it would # eventually rewrite a correct sentence into a wrong one. Common words come out wrong because the audio # is poor, and that is fixed at the recorder (persist.helpdesk.rec_bitrate), not here. # # Before adding a term, ask: could a normal Czech word be one or two edits away from it? If yes, leave # it out. Every entry is applied to every call. # # Verify a change with: ruby components/backend/test/test_glossary.rb # --- systems -------------------------------------------------------------------------------------- yamt # the monitoring platform; Whisper writes "jamt" - same sound, no way to know repeater plánovač RF plánovač # --- operators ------------------------------------------------------------------------------------ Vodafone T-Mobile # --- sites ---------------------------------------------------------------------------------------- # Add site names as they come up in real calls. These are exactly the words that matter to a technician # and exactly the ones a general-purpose model has never seen. Ježlovice