Leor | 2024-03-20 13:21:22 UTC | #1
Hi community!
I'm Leor - one of the product manager's looking after voice transcription in Genesys Cloud.
We recently launched Dictionary Management through API which allows for accuracy improvement in native voice transcription.
You can find the documentation in API explorer here: https://developer.genesys.cloud/analyticsdatamanagement/speechtextanalytics/#delete-api-v2-speechandtextanalytics-dictionaryfeedback--dictionaryFeedbackId-
The API allows for several inputs to the dictionary:
dialect term examplePhrases phrases source boostValue soundsLike
At a basic level, dictionary management (referred to in the API as dictionaryfeedback) allows developers to add commonly mis-transcribed words so that they can be better detected by voice transcription. This can be particularly useful for business or industry specific terminology, company or product names.
A little explanation on the variables to get you started:
dialect – this is to select the dialect from native voice transcription for which we’ll be improving the transcription term – this is the term spelled exactly as it should appear in the transcription. It should not contain special characters or spaces. examplePhrases – these are example phrases in which the term usually appears source – for the time being, “manual” is the only supported source, as in we are manually added words to dictionary. Future features will support more capabilities of added terms boostValue – this is the amount for which we increase the likelihood of the transcription service identifying the term. It follows a log scale from 1 to 10.
Here’s an example:
{ "dialect": "en-US", "boostValue": 3.5, "term": "acme", "examplePhrases": [ { "phrase": "I like acme brand pancakes", "source": "Manual" }, { "source": "Manual", "phrase": "the brand acme is my favorite" }, { "phrase": "loyal acme customer" } ], "soundsLike": [ "acne" ] }
The limits to fields are defined in API limits here: https://developer.genesys.cloud/organization/organization/limits
Please note that transcription does not yet support the normalization of digits, so numbers like "2" will show up as "two" and dictionary management will ignore terms that include digits. We hope to have this updated later in 2024.
Would love your feedback! We hope to have UI access in a few months.
Sincerely,
Leor leor.grebler@genesys.com
system | 2023-11-16 17:31:08 UTC | #2
This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.
This post was migrated from the old Developer Forum.
ref: 22563