Word Boosting

View as MarkdownOpen in Claude
Real-Time

Word Boosting (also known as Keyword/Keyterm Prompting) allows you to improve Keyword Recall Rate (KRR) for important keywords or phrases by up to 90%. This feature helps ensure accurate transcription of industry-specific terminology, product names, company names, and specialized jargon.

Enabling Word Boosting

Add keywords parameter as an array of strings in the format word:weight where weight is a positive number indicating boost intensity.

Word Boosting is currently only available for the Real-Time WebSocket API.

Real-Time WebSocket API

1const url = new URL("wss://waves-api.smallest.ai/api/v1/pulse/get_text");
2url.searchParams.append("language", "en");
3url.searchParams.append("encoding", "linear16");
4url.searchParams.append("sample_rate", "16000");
5url.searchParams.append("keywords", JSON.stringify(["nacho:5.0", "bacon cheeseburger:4.0"]));
6
7const ws = new WebSocket(url.toString(), {
8 headers: {
9 Authorization: `Bearer ${API_KEY}`,
10 },
11});

Case Sensitivity and Formatting

Keywords preserve formatting (including case and punctuation) which helps control how proper nouns, product names, or company names are transcribed.

Best practices for keyword formatting:

  • For proper nouns (names, brands, titles): Use appropriate capitalization (Waves, iPhone, Dr. Smith)
  • For non-proper nouns: Use lowercase (tretinoin, algorithm, protocol)

When smart formatting is applied to the transcript, words that start sentences may be automatically capitalized regardless of keyword formatting.

Expected Improvements

Word Boosting can significantly improve recognition accuracy and confidence scores for industry-specific terminology. Typical improvements include:

SourceBefore RecognitionAfter Recognition
nacho stack double crunch taco”macho stack""nacho stack double crunch taco”
bacon cheeseburger”bake in cheeseburger""bacon cheeseburger”
account number”a count number""account number”
billing department”building department""billing department”
technical support”tech nil call support""technical support”

Best Practices for Keyword Selection

Medical terms (tretinoin:5.0, diagnosis:4.0), technical jargon (escalation:4.5, API:3.5)

Brand names (Waves:5.0, iPhone:5.0), service names, competitor names

Common phrases in your domain (account number:4.0, customer service:3.5)

Names, brands, titles with appropriate capitalization (Dr. Smith:5.0)

Use lowercase (algorithm:3.0, protocol:3.0, refill:4.0)

What to Avoid

Very common words that are rarely misrecognized (the, and, is) - don’t boost these

Words that appear in many contexts without specific meaning

Focus on the most important 20-50 terms with appropriate weight values

Ensure capitalization matches your desired output

Avoid using only very high weights (10+) or very low weights (1) - use a range (1-10) to differentiate importance levels