AI Evaluation

Paramoji Speech Flow

I love crafting and making speeches.

I love tools which analyse and give insights into my speeches.

Last week I attended a AI Innovators meetup organized at KI Reply in Munich. One of the sessions was “Human emotions at scale” presented by Stefan Dirnstorfer.

I liked the way he compared the tweets and opinions of Trump and Biden before and after the debate, Rishi and Keir before and after the election results using Paramojis.

A paramoji is a parametric smiley face that represents facial expressions on continuous scales” – Stefan Dirnstorfer

I was wondering if I could follow the emotional rollcoaster that a speech takes the audience through.

Using ChatGPT, GitHub Copilot and the PHP code snippet from Stefan I wrote a small application which created the above Paramoji Speech Flow.

I like the way in Page 1, you can note that emotionally in every 5 sentences there is happy and sad emotion (deliberately simplifying emotions for discussion)

I like the way in Page 4, you can notice the excitement in the mouth and eye gestures.

How do you analyse and learn from insights from your speeches?


[HYBRID] AI in Software development lifecycle, Wed, Jul 10, 2024, 6:00 PM | Meetup



GPT4All for Insights into speeches

A lot of speakers have made speeches over the years and a have quite a collection of them.

What are we doing with these speeches?

Are we mining them for insights into the messages that they had?

Are we dissecting to find common messages that we are comfortable and good at talking about?

I played around with GPT4All with my district level speeches and tried to understand the message from one speech and the common message across multiple speeches.

Check out the YouTube video and see if you would try something similar with your speeches?

Link: GPT4All