Quality, measured
Precise with statutory sections, amounts and deadlines. Measured, not promised.
A wrong filler word hardly matters. A wrong section number or a lost “not” does. That is why we measure exactly these spots. In a comparison of eight speech recognition systems, the one in Simpliant Dictation is in the leading group.1
How we measured
Thirteen German legal dictations – extracts from the Civil Code, Criminal Code, GDPR and Commercial Code, two Federal Court of Justice decisions and four fictitious law-firm dictations – were spoken by four synthetic voices, 15:43 minutes in total. Each of the eight speech recognition systems received each file once, in German and without a dictionary. We evaluated the word error rate and 82 check points: sections, amounts, dates, case numbers and negations. Only speech recognition was measured, not the app.