Skip to content
This repository was archived by the owner on Dec 3, 2022. It is now read-only.

salcc/Junction-2022

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

105 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Junction 2022

What have we built?

We have built a data-driven platform that allows users to:

  • Express and track their mental status
  • Anonymously connect and empathize with other users based on likelihood of improvement through a chat channel.

Solution contents:

  • Flutter working demo app with the basic features of our platform.
  • Matchmaking model based on a large pretrained embedding transformer mapping documents to latent space representations gathering similarities and user profile extracted features. The model is re-trained through a metric lerning approach based on a custom adaptation of a contrastive loss which considers user feedback and status tracking data.
  • Chat assistant bot modeled as an agent in a chat enironment where conversation score-based rewards and penalizations are extracted to apply reinforcement learning.
  • Sample Dataset containing a definition and some rows of a lightweight pseudo-datawharehouse.

Key aspect, automated improvement of the solution based on the usage gathered data unlocking:

  • Re-training our matching model based on user feedback and improvement results
  • Conversation data defining an environment for a virtual chat assistant dealing with ice-breakers, destructive conversations,etc. aiming to maximize connection quality.

The more usage the app has the better performance the AI could bring.

Videos:

Authors:

Jordi Baroja
Marçal Comajoan
Mauro Filomeno
Marc Franquesa
Pol Puigdemont

About

Junction 2022 Hackathon

Resources

Stars

Watchers

Forks

Contributors