SpeechCoach.ai

homepage
results

Inspiration

I have always been scared of speaking in public and used to ask my sister's help for live feedback about my speaking skills. In the corporate world there is always a necessity to improve your public speaking skills, and we thought why not build an AI powered application that helps you do so

What it does

SpeechCoach.Ai listens to you speak, compares your speeches to some of the best speakers in the world like Obama and gives you immediate feedback. The feedback includes features like your pace of talking, how many filler words you used, sentiment analysis and number of grammatical mistakes

How I built it

We used the IBM Bluemix API which converted the audio output recoded to text. The text included information about time taken to speak each word as well as confidence. With these inputs we came up with heuristics to determine the the features - pace of words, filler words. We used language tools to compute sentiment analysis and grammatical mistakes

Challenges I ran into

Not all speeches available online are suitable data for testing. We had to manually crop the available data to train our models. Also, all of us are backend engineers with little or no front end experience and we ran into few issues while building our web application.

Accomplishments that I'm proud of

I am really proud of pulling our whole idea together in less than 24 hours and building a user friendly web application. We learnt a lot about the different research in speech detection, language processing and in general how to get things working quickly (Crushing things)

What I learned

sentiment analysis, speech detection, language processing

What's next for SpeechCoach.ai

We plan to integrate with video to detect facial expressions and compute relevant feedback and live stream the feedback. We will also be conducting user research surveys to gauge how easy it is to use our product it.

Built With

Updates