Integration with automation tools enhances customer sentiment analysis
What is our primary use case?
I have used Amazon Comprehend primarily for sentiment analysis in my project. I analyze customer transcripts to determine if they are satisfied with the agents they interact with. I store the transcripts in DynamoDB and pass them sentence by sentence to Comprehend to get a score indicating positive, negative, or neutral sentiment. For instance, if there are ten consecutive negative sentiment scores, it suggests the customer is unhappy with the agent. This sentiment analysis is conducted using AWS Lambda, which calls the Comprehend API.
What is most valuable?
Comprehend is a useful service for sentiment analysis as it analyzes customer transcripts to evaluate interactions between customers and agents. It provides scores indicating whether sentiments are positive, negative, or neutral. The integration with AWS services like DynamoDB and Lambda facilitates automated analysis, contributing to more informed assessments of customer interactions.
What needs improvement?
Regarding improvements, I would focus on accuracy. For example, if a customer says, 'I want to cancel my credit card,' it should clearly be identified as a negative sentiment. Improving accuracy in such detections would yield more reliable results.
For how long have I used the solution?
I have used AWS Amplify for almost four or five years. As for Amazon Comprehend, I have used it a lot recently, especially in my current project.
What do I think about the stability of the solution?
Amazon Comprehend is a stable product and has been around for quite some time.
What do I think about the scalability of the solution?
In terms of scalability, Amazon Comprehend is scalable; however, as it is an ML-based service providing sentiment scores, scalability considerations are less critical compared to server and database scaling.
How are customer service and support?
I am totally happy with Amazon Comprehend support. AWS support provides excellent follow-ups.
How would you rate customer service and support?
Neutral
How was the initial setup?
The initial setup for Amazon Comprehend is completely easy. As an AWS service, there isn't much to deploy. Using a CloudFormation template for serverless deployment is quite straightforward.
What's my experience with pricing, setup cost, and licensing?
Amazon Comprehend is somewhat costly. While it is not prohibitively expensive, the costs can add up, especially in comparison to self-written NLP models. However, it is worth it if you lack developers with expertise in writing such models.
What other advice do I have?
I would rate Amazon Comprehend an eight out of ten because there is always room for improvement, especially in terms of accuracy. For those new to Comprehend, understanding its usage and reviewing the AWS documentation and instructional videos is essential. Additionally, evaluating whether developing custom NLP models could be more cost-effective is crucial.
Which deployment model are you using for this solution?
Public Cloud
If public cloud, private cloud, or hybrid cloud, which cloud provider do you use?
Amazon Web Services (AWS)
