My friend and I were on the same flight, which was delayed by one hour, with horrible onboard service. After getting off the flight, we were sharing a cab back home - when we both got notifications to rate our flight journey. I gave it one star and didn't write anything about why I did so. My friend, on the other hand, gave it a 3 star. I was intrigued, and I asked him why not 1. He told me the lowest rating he gives to products or services is 3. I was intrigued to see this.

Collecting Feedback is common across nearly all e-commerce businesses, be it a products company like - Amazon or a Services company like - Airbnb.

What it does

Talk Back SDK allows businesses to add the ability to collect Voice-based feedback from their end-users. This can be added to any feedback collection screen or page. The SDK, on addition, shows up a mic button on the feedback collection screen and prompts the user to use voice to give her feedback. We transcribe this text using streaming ASR and then run sentiment analysis on the text using the API offered by This allows us to give each user a better and consistent rating based on the words they used while giving feedback.

How we built it

We have used a demo e-commerce APK to showcase a proof of concept where we ask the user to give feedback on her latest purchase. The user has to give input using voice. We use streaming ASR from a third party to transcribe the speech. We then use state of the art, sentiment analysis API provided by to run sentiment analysis on the text to provide a rating to the review.

We collect long-form qualitative data and give a consistent quantitative rating to that review, leading to more clarity on the customer's voice. Quite literally.

Challenges we ran into

Some of the challenges we faced were:

  1. Transcription of the text needed to be accurate.
  2. Language issues associated with this.
  3. Privacy concerns that users might have.

Accomplishments that we're proud of

One of the biggest accomplishments - We didn't even think about this before building the TalkBack SDK was - During testing, I gave the demo App to my grandmother and told her that we need her to give her feedback for the sewing machine (I recently purchased for her on Amazon) - Seeing my grandmother express her appreciation for the sewing machine easily with Voice was truly something magical. Enabling accessibility was one of the use cases we didn't think of before venturing into creating this.

What we learned

There are more than 20 million e-commerce websites on the internet, and more are coming up daily. Almost all of these e-commerce websites have one thing in common, collecting feedback. Instead of making it for one app, we decided to build TalkBack SDK as a product that can be added to any feedback screen on any e-commerce applications, providing the ability for e-commerce businesses to collect rich qualitative feedback based on quantitative sentiment analysis.

What's next for TalkBack SDK

We want to use's NLP APIs to provide even richer analysis on the feedback and bucketize them based on different entities and criteria to give an even clearer snapshot of the reviews to businesses. Effectively providing a more clear voice of the customer inside every business.

Built With

Share this project: