How do I improve accuracy with the Speech Anywhere Widget?

Overview

This article covers the steps you can take to maximise transcription accuracy when using the Speech Anywhere Widget.

Applies to

All Users, Administrators

Use a quality microphone

The single most important factor in accuracy is microphone quality. A dedicated USB headset microphone or professional speech microphone (such as a Philips SpeechMike) will deliver significantly better results than a built-in laptop microphone. Ensure the microphone is positioned correctly typically 1 to 2 inches from the corner of your mouth, slightly to the side to avoid breath sounds.

Reduce background noise

Dictate in a quiet environment where possible. Background noise such as air conditioning, office chatter, or music can affect recognition accuracy. If you must dictate in a noisy environment, consider a noise cancelling microphone.

Speak clearly and naturally

Speak at a natural pace not too fast and not too slow. Do not exaggerate pronunciation. Speak in complete sentences where possible, as context helps the AI engine choose the correct words.

Use speech to text replacements

Use the Speech to Text Replacement feature in your Dictalogic account to teach the system your preferred spellings, proper nouns, abbreviations, and industry specific terminology. For example, if the system consistently misrecognises a client name or a medical term, add it as a replacement. See the knowledge base article “How to add a Speech to Text Replacement?” for guidance.

Train the widget to your voice

Speech Anywhere’s AI engine learns from your usage over time. The more you use it, the better it becomes at recognising your accent, speech patterns, and vocabulary. Consistently correcting errors rather than leaving them helps the system improve faster.

Dictate punctuation

Always dictate punctuation aloud. This significantly improves the readability of transcribed text and reduces post dictation editing time.

Check your internet connection

Speech Anywhere requires a stable internet connection to send audio to and receive transcription from the Microsoft Azure cloud. A slow or intermittent connection can lead to delays or errors. Use a wired connection or a strong Wi-Fi connection where possible.

Keep the widget updated

Ensure you are running the latest version of the Speech Anywhere Widget. Updates often include accuracy improvements and new features.