Social Rescue Tech example

Positive NLG Compliments

./compliments.txt

Demonstrations

Video demonstration

https://asciinema.org/a/w5ayuAEGGpgsxTCKWbAGRPfvZ

Video demonstration (mp4)

./demo1.mp4

Video demonstration (gif)

8000+ and still generating

https://asciinema.org/a/YexeaMmvc15iaGzHQFLMD1tr0

Video demonstration (mp4)

./demo2.mp4

Video demonstration (gif)

And still generating

https://asciinema.org/a/cuCzjyq5EwT6dV8Xp3RIk9B3l

Video demonstration (mp4)

./demo3.mp4

Video demonstration (gif)

Code

Prompt

http://github.com/semiosis/prompts/blob/master/prompts/compliment-generation-based-on-a-celebrity-1.prompt

Final generated prompt:

The following are nice compliments:
###
TOPIC: Jason Momoa
- He has fame, but he doesn't want it. He has advantage but he doesn't take it. And he definitely has all our hearts!!!!!
###
TOPIC: Amber Heard
-

Simulate real conversation using NLU and NLG

Demonstration

https://semiosis.github.io/apostrophe/

Conversation demo 1

https://asciinema.org/a/K40px4H4CPPN15QMz6Uy8Pz3q

Video demonstration (mp4)

./conv-demo1.mp4

Video demonstration (gif)

Conversation demo 2

https://asciinema.org/a/aLn5lZq5yIeBxR6dyFyveDJGc

Video demonstration (mp4)

./conv-demo2.mp4

Video demonstration (gif)

Conversation demo 3

https://asciinema.org/a/YdX4bzPownHsVT3T0UPLMcp3c

Video demonstration (mp4)

./conv-demo3.mp4

Video demonstration (gif)

Time taken to generate

git show -s --format=%at a9fccf1

git show -s --format=%at 08b7c56

# Seconds
echo "1639635963 - 1639627967" | bc

# Total minutes
echo "(1639635963 - 1639627967) / 60" | bc

It took just over 2 hours to generate 10k compliments.

This pipeline will be sped up in the near future by a factor of around 10x.