After a long day of editing English essays using Grammarly, we thought to ourselves: Why isn't there a similar service for presentations? That was when Speech Teech was born.

What it does

Speech Teech is a website that analyzes recordings of speeches using Azure API's, showing statistics such as WPM, confidence level, and filler words to the user. Afterwards, it also shows recommendations for the user on how to improve his/her speaking.

How we built it

Using node.js and express, we created a backend that would send recorded speeches to text and speech analytic API's developed by Azure. We then developed a web app to combine the analytical results and wrote additional code to make the website aesthetically pleasing. We also developed a machine learning model based off of a data sample of >2500 Ted Talks to predict how engaging it is.

Challenges we ran into

  • We had trouble learning how to make the backend and frontend talk to each other. This was our first time using node and react, and we learned a lot from stack overflow in order to solve these challenges over the nights.
  • Working together in a git environment as this was all our first time collaborating. (also we had a teammate who snores loudly)
  • We also had trouble adapting to javascript syntax as we were new to coding in js. Turns out debugging why System.out.println("hello world"); didn't work was a bad idea at 3:00 am in the morning.

Accomplishments that we're proud of

We were able to sleep more than an hour a day while still being able to type up ~clean~ code. But more importantly, we collaborated using a good workflow, and ended up with a product we could use in our daily lives.

What we learned

  • We learned how to utilize node and express, and also got a hang of REST API's.
  • How to use MIcrosoft Azure's API's and machine learning studio.
  • Coffee is overpowered

What's next for Speech Teech

In the future, we want a better user experience such as an easier way to record and transfer their audio. We would also like to collect and train a data set of filler words to enhance our algorithm. We would also make more use of API's outside of Azure in order to incorporate more analysis of the user's speaking in order to increase the amount of help we can give them. By getting a better hold of HTML and CSS, we also intend on increasing the interactivity of our website by including popups and graphs.

Share this project: