How accurate is Speech Anywhere compared to human transcription?

Overview

This article explains the accuracy levels achievable with Speech Anywhere and how it compares to human transcription.

Applies to

All Users

AI accuracy

Speech Anywhere uses Microsoft Azure Cognitive Speech Services, a state of the art AI engine trained on industry specific vocabularies. Dictalogic’s platform is designed to achieve up to 99.8% accuracy under optimal conditions, which is comparable to experienced human transcription.

Factors that affect accuracy

Accuracy depends on several factors including microphone quality, background noise levels, the user’s accent and speech clarity, the use of specialist terminology, and whether Speech to Text Replacements have been configured for domain specific terms. New users may experience slightly lower accuracy initially, which improves as the system adapts to their voice.

Compared to human transcription

For real-time, live dictation, Speech Anywhere provides a significant speed advantage over human transcription. A human transcriptionist typically processes audio at a ratio of 3:1 to 4:1 (i.e., one minute of audio takes 3–4 minutes to type). Speech Anywhere delivers results instantaneously. For highly complex, accented, or technical dictations, human transcription via the Dictalogic workflow may still offer an advantage in accuracy. Many organisations use both Speech Anywhere for everyday correspondence and quick notes, and the cloud dictation workflow with human transcribers for complex formal documents.