DEV Community

Cover image for Automated Customer Care Rating
Amogh Singh
Amogh Singh

Posted on

Automated Customer Care Rating

Innovative Ideas Challenge

Introduction

The Innovative Ideas category was an ideal choice for me, as I was interested in pitching a project that would be useful and allow for quicker results and more accurately. Before taking part in Deepgram Hackathon, I have had a bit of prior experience with Speech Recognition Technology, mainly out of curiosity and exploration.

My Deepgram Use-Case

My inspiration for this submission comes from customer care centers, where the executives are often helpful but there are cases where executives are rude to their customers, which has a negative impact larger than the positive impact made. A single bad experience can spoil a customers relationship with the organization/company, and in the long run can be harmful to business. Deepgram can be used here to generate transcripts of the call recordings usually stored by the customer care centers and appropriately evaluate a executive's efficiency and their talking style with customers.

Upon exploration of the Deepgram website and their API documentation, I found a few use cases of call transcription and realized this would be an ideal use-case for Deepgram, and therefore chose this particular topic.

Dive into Details

Customers are the utmost priority for any company or organization, and they must be treated so, especially when they seek help from a customer care. But there are cases where customer care executives are rude to a customer.

This is where Deepgram comes into picture. Deepgram can be used to convert those call center recordings to text, which can then be processed by a Machine Learning model, to generate the general
emotion of the customer care executive through the words they chose, and the emotion of the reply given by the customer. This can then be used to score and better train the executives to handle customers well and allow for a better service to the end customer.

This method takes into consideration the emotion or attitude of the customer as well as the executive to better judge the executive and can be utilized appropriately to eliminate cases where the customers are being rude with no fault of the executive.

Conclusion

Having gone through the Deepgram documentation, it can certainly achieve the task of call transcription, and coupled with Machine Learning models for emotion detection, it could be used for efficient handling of customers and training of customer care executives.

The idea of speech recognition, and the extraction of emotion through text, is the two main components of this idea, with both of them being highly researched, and now, widely used technologies, for various use cases ranging from hate speech detection to captions generation.

Through the course of research for this project, I feel that with the combination of these two technologies, there are many more applications which can be devised and used to automate simple tasks like the one discussed.

Top comments (0)