Can live transcription be made more accurate if I know the specific words I'm listening for from the user? I just want to very accurately verify that the user has said specific words.
#Can live transcription be made more accurate if I know the specific words I'm listening for?
remote geyser
ruby swiftBOT
Hi @remote geyser. I'm kapa.ai, your friendly support bot. I will try to answer your question by searching through the knowledge base. :wave:
Yes, Deepgram provides features that can help improve the accuracy of live transcription, especially if you know the specific words you're listening for.
One of these features is the Keyword Boosting feature. This feature allows you to send along with the audio, words that Deepgram should watch more closely for, so it can transcribe them more accurately. You can send up to 200 keywords per API request. If you want a keyword capitalized in your transcript, capitalize it in the keywords parameter. This feature is useful when you want to improve the accuracy of specific keywords or vocabulary source (https://developers.deepgram.com/docs/migrating-from-amazon-web-services-aws-transcribe-to-deepgram#custom-vocabulary).
Another feature that can be useful in your case is the Search feature. Deepgram's search is based on the acoustic sounds of keywords, allowing you to accurately identify whether a phrase was uttered in submitted audio. You can pass in a word or phrase to find and then get results in the response JSON object, including the start time and end time of when that phrase was possibly uttered and a confidence rating for each match. You can include up to 25 search terms per request source (https://developers.deepgram.com/docs/migrating-from-amazon-web-services-aws-transcribe-to-deepgram#search).
However, it's important to note that the effectiveness of the Keywords feature can depend on the quality of Deepgram's off-the-shelf models. The better these models get at transcribing on their own, the less powerful the Keywords feature becomes because Deepgram gives an increasing confidence level to the transcription itself. If you are using an extremely powerful off-the-shelf model, using keywords might not have an effect source (https://developers.deepgram.com/docs/keywords-vs-search#demos).
Therefore, it's recommended to benchmark the transcription results without any keywords, and then add words one by one to notice the effect. Until you get a feel for how the Keywords feature works with your chosen use case models and unique language context, you may find that providing keywords seems to have no effect or even lower the quality of the transcription source (https://developers.deepgram.com/docs/keywords-vs-search#demos).
Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:
remote geyser
Does using Keyword Boosting work with live transcription?