loading

Business Impact

  • Post Call Analysis

  • Ensure Scalability

  • Improve Customer Experience

Customer Key Facts

  • Location : India
  • Industry : Insurance

Problem Context

The client has live agents attending calls to address customer queries related to insurance services and products. They follow a script to welcome customers and introduce any new product/request. The client wanted to perform post-call analytics to analyze customers’ sentiment and few metrics to analyze the customer experience. Therefore, they built a solution to transcribe the audio and perform sentiment analysis based on voice. However, there is a requirement to improve this solution as it is unable to perform well with mono format due to the presence of multiple languages in single audio and Indian English.

Challenges

 

  • Building a specific solution for regional English dialect with abundant usage of native language
  • Lack of labelled data and thereby ensuring the quality of data tagging
  • High diarization accuracies for mono format calls

Technologies

Amazon Transcribe
Amazon SageMaker
Amazon Quicksight
AWS Lambda
Amazon S3
Amazon Translate
Amazon ECR
Amazon Athena
Amazon Glue

Building a Customized Model to Transcribe the Audio and Perform Sentiment Analysis Based on Voice

Solution

Quantiphi has proposed the following solution to help the customer with:


Transcribe: Natural Language Processing-based solution using Amazon Transcribe which will be able to translate the call audio to text and segregate as per the speakers with the help of custom vocabulary.


Custom Model for Sentiment Analysis: The output from the above along with the audio will be used to determine the sentiment of the conversation by using a hybrid model i.e., using both text-based and acoustic-based approach.


Dashboards: Develop dashboards in QuickSight on a predefined framework of fixed data / schema provided by customer. This will help to better understand the insights and perform call center analytics.

Result

  • Standardize and setup the process of evaluation for customer satisfaction and agent performance
  • Recognize what has been done and what needs to be improved at an agent level by displaying sentiment and word-level and overall statistics
  • The solution is scalable to n number of customers in order to provide a better customer satisfaction
  • Enable reporting to take strategic decisions based on the data via dashboards

Looking for similar project?

Let's Talk

Get your digital transformation started

Let's Talk