Inspiration

We found that there was no effective automated tool to evaluate presentations. When I sample a presentation to someone I know, they can be biased and not highlight all the negative points of my presentation - or they might miss critical advice that would've been useful to improve. So, we set out to build a presentation kit that can critique all parts of the presentation - posture, speed, tone, and overall impression given to the audience. This way, presenters can get a consistent and reliable review of their presentation.

What it does & How it's made

This AI powered tool does exactly that - it gives concrete and unbiased advice on how to improve a presentation. PresentationKit is made using node.js and managed using yarn. The backend employs Tensorflow, specifically the MobileNet model to capture the body structure of the presenter. This helps yield a score for the posture of the presenter over the entire presentation. The audio from the speech is streamed to Google's Speech-to-Text API, which returns the tokenized words spoken in real-time. Using this data, we track how many words are spoken per minute and how many filler words are used. Additionally, we pass the text through sentiment analysis on Google to figure out how positive or negative the presentation is - the presenter could use that information to adjust their word choice and make a presentation more positive, for example.

Challenges

There were several challenges - most centered around integrating all the components in Google Cloud and Tensorflow into node.js.

Accomplishments

I think it is an accomplishment to have finished what we set out to do - to begin, there were vague requirements about what made a presentation good or successful. After finding out how to determine a successful presentation, we had to research technologies that could yield insights about the presenter's body language and content.

What we learned

We learned that it is pretty tricky to iterate through an entire project in only 24 hours! Not that by any means it is finished, but it's great to have a working product that can actually add value for people who want to be better presenters.

The future of PresentationKit

While the services PresentationKit offers are really valuable, the UI that presents them to the user could be more user friendly. One aspect would be to invest time improving the UI. Another would be to create a proper platform, where users could log in and out and check their history of presentation trials. Hopefully they would see improvement, both on the platform and in the real world!

We could refine the services by trying to examine the inflection in the presenter's voice - that can also heavily influence the sentiment of a presentation.

Also, the impact of the presentation is also determined so much by the content - if they leave the audience with a good question to think about, that makes a presentation more effective. So there is likely a lot more that could go into the platform to help presenters improve.

Built With

Share this project:

Updates