Med-1 is Omnimedica’s proprietary speech-recognition engine for Greek clinical language, the engine that turns a doctor’s dictation into precise, reliable text.
Med-1 is Omnimedica’s proprietary speech-recognition engine for Greek clinical language, the engine that turns a doctor’s dictation into precise, reliable text.
Greek is hard. Medical Greek is harder. Off-the-shelf speech-to-text trips over clinical terminology, abbreviations, declensions and the way doctors actually dictate, and one wrong word in a report is one too many.
Where general speech-to-text slips
General STTμικρά οστά φυτά στο γόνατο
Med-1μικρά οστεόφυτα στο γόνατο
The arithmetic of learning
The words a clinic loses are always the same words.
We compared 213 signed reports against the raw transcripts of the same dictations. The distance between the two bars is the number of times a radiologist retyped the same word by hand.
in the signed reportin the raw transcript
exiting nerve roots
8530
+55 manual corrections
mass-like lesion
5415
+39 manual corrections
paraspinal space
5012
+38 manual corrections
acromion
232
+21 manual corrections
The size of that gap sets the entry’s recognition weight, from 0 to 20. The more often a word was lost, the harder the system now listens for it.
Audit of 213 completed exams in a live deployment.
What Med-1 does
Tuned for the language of the clinic.
Greek-first, not translated
Med-1 is built for Greek as it’s actually spoken in clinics, declensions, word order and phrasing included.
A medical vocabulary
Anatomy, pathology, procedures and drug names that general speech-to-text mishears or smooths over.
Abbreviations & dictation habits
The shorthand and conventions doctors actually use when they dictate, expanded correctly.
Accent-insensitive understanding
Matches names and terms regardless of accentuation, so ΜΑΡΙΑ, Μαρία and μαρια are the same person.
Patient-name detection
Med-1 surfaces the patient’s name from the dictation so it can be confirmed before anything is finalised.
Two passes, no waiting
A live word-by-word preview while you speak, then an authoritative transcription the instant you stop.
