Foreign words / names in custom vocabulary for Amazon Transcribe now that SoundsLike is deprecated

Since the SoundsLike and IPA columns are no longer supported for Custom Vocabulary in Amazon Transcribe, you'll need to adapt your approach for foreign names and words.

The best current method is to focus on the Phrase and DisplayAs columns:

1. In the Phrase column:

• Format multi-word terms with hyphens (e.g., "Alphonse-Mucha")
• Don't use spaces
• Spell out numbers rather than using digits

2. In the DisplayAs column:

• Define how you want the term to appear in your transcription output
• You can use spaces here (e.g., "Alphonse Mucha")
• You can include digits if needed

For your examples, your custom vocabulary table would now look like:

Phrase DisplayAs chamois chamois Alphonse-Mucha Alphonse Mucha philip-de-laszlo Philip de László

While you can no longer provide pronunciation hints through SoundsLike, Amazon Transcribe will still attempt to recognize these terms based on the Phrase column. For better accuracy with domain-specific terminology or foreign words, you might want to consider using a custom language model in addition to your custom vocabulary, as it can provide more context-aware recognition by learning from a large volume of domain-specific text.
Sources
Creating a custom vocabulary using a table - Amazon Transcribe
Improving transcription accuracy with custom vocabularies and custom language models - Amazon Transcribe