Yet Another Tone Analyser - YATA

Inspiration

Likewise R&D Opportunity + Halloween weekend fun!

What it does

We intended to build a platform that emits customer emotion & facilitate necessary action based on their speech.

How we built it

We build it using

• Kinesis for Data Stream
• Flink for Stream Processing
• ParallelDots for Emotional Analysis
• webkitSpeechRecognition for Speech to Text Conversion

Challenges we ran into

• We were new to Google NLP modules, causing technical hiccups here and there. So we shifted to webkitSpeechRecognition.
• Emotion Analysis using AutoML also posed some challenges. So we shifted to ParalleDots.

Accomplishments that we're proud of

• E2E flow was completed just a few hours before the end.
• Project Naming
• Project Structure
• Tech choices

What we learned

• Google API's are something to be explored before a hackathon
• Some things need a simpler solution

What's next for YATA

Scaling the code for a large user base.