Google Sentiment Analytics

Summary

In this post, I'll demonstrate use of a Google Natural Language API - Sentiment.  The scenario will be a caller's interaction with a contact center agent is recorded and then sent through Google Speech to Text followed by Sentiment analysis.

Implementation

The diagram below depicts the overall implementation.  Calls enters an ACD/recording platform that has API access in and out.  Recordings are then sent into the Google API's for processing.

Below is a detailed flow of the interactions between the ACD platform and the Google API's.

Code Snippet

App Server

Simple Node.js server below representing the App Server component.

app.post(properties.path, jsonParser, (request, response) => {
 //send a response back to ACD immediately to release script-side resources
 response.status(200).end();
 
 const contactId = request.body.contactId;
 const fileName = request.body.fileName;
 let audio;
 
 logger.info(`contactId: ${contactId} webserver - fileName:${fileName}`);
 admin.get(contactId, fileName) //Fetch the audio file (Base64-encoded) from ACD
 .then((json) => {
  audio = json.file;
  return sentiment.process(contactId, audio);  //Get transcript and sentiment of audio bytes
 })
 .then((json) => { //Upload the audio, transcript, and sentiment to Google Cloud Storage
  return storage.upload(contactId, Buffer.from(audio, 'base64'), json.transcript, JSON.stringify(json.sentiment));
 })
 .then(() => {
  admin.remove(contactId, fileName); //Delete the audio file from ACD
 })
 .catch((err) => {
  logger.error(`contactId:${contactId} webserver - ${err}`);
 });
});
 
app.listen(properties.listenPort);
logger.info(`webserver - started on port ${properties.listenPort}`);

Source: https://github.com/joeywhelan/sentiment

Copyright ©1993-2024 Joey E Whelan, All rights reserved.