Google Speech-to-Text Transcriber
A plugin for IVR Tester to transcribe IVR call flows using Google Speech-to-Text.
Installation
1. Install the package
Install using yarn:
yarn global add ivr-tester-transcriber-google-speech-to-text
Or npm:
npm install -g ivr-tester-transcriber-google-speech-to-text
2. Setup Google Speech-to-Text
Follow Google Cloud Platform’s quick-start guide to:
- Setup a Google Cloud project
- Enable the Google Speech-to-Text service
- Create a service account
- Create then download a JSON key-file for the service account
- Set the
GOOGLE_APPLICATION_CREDENTIALS
environment variable to the path of the JSON key-file
export GOOGLE_APPLICATION_CREDENTIALS=/path/to/key.json
Configuration
languageCode
- Language of the speech to transcribe - list of available languagesuseEnhanced
- Whether to use an enhanced model if it is available for the language code provided. Beware this costs more!speechPhrases
- Specific words or phrases used more frequently within the speech. See Google’s guide for more info on what to provide
Usage
CLI
Create the following block within your configuration file, adjusting the options
property based on your needs:
{
"transcriber": {
"name": "google-speech-to-text",
"options": {
"languageCode": "en-GB",
"useEnhanced": true,
"speechPhrases": [
"Press 1 for playback flow",
"Press 2 for long pauses flow",
"Press 3 for short latency flow",
"Press 4 for long latency flow",
"Please enter a number",
"You entered the values 0123456789. Thank you."
]
}
}
}
Programmable interface
import { googleSpeechToText } from "ivr-tester-transcriber-google-speech-to-text";
const config: Config = {
transcriber: googleSpeechToText(
{
languageCode: "en-GB",
useEnhanced: true,
speechPhrases: ["balance", "costing", "financial services"]
}
)
};
See the API’s documentation for details of each parameter.
Development
The code is part of IVR Tester’s GitHub repository.
Debugging
Console logging can be enabled by specifying the package name in the DEBUG
environment variable:
DEBUG=ivr-tester-transcriber-google-speech-to-text
DEBUG_DEPTH=5