Keyword boosting allows users to instruct a model to pay extra attention to specific words. List the word(s) in your request configuration if using an SDK or in your query parameters if making an HTTP request. By default, the feature boosts the listed word, but you can apply a number to enhance (positive) or suppress (negative) the word.

Keyword boosting works best for single word proper nouns, especially if Deepgram’s model does not output the word correctly. For example, if “Deepgram” is not recognized properly, boost it in your request.

Note: Phrases, acronyms, digits, and special characters are not supported by Deepgram’s Keywords feature. For instance, “web 3.0” will be separated into “web” and “3.0”, boosting each term separately. Boosting common words like “web” can result in false positives, and terms like “3.0” won't work due to numerals and special characters.

Keyword Boosting vs. Custom Model Training

If you spend too much time configuring the keyword feature, consider a custom-trained model to learn and identify your key terms. Use a custom model if:

• The number of keywords is excessive
• Many words are outside your custom model’s vocabulary
• You need to emphasize terms that are not single proper nouns

Conclusion

The keyword feature offers light guidance for your model. If your configuration becomes too prescriptive, consider training a custom model.

For more information on custom model training, see the product page or contact a Product Consultant: https://deepgram.com/contact-us/