This shiny-app repository is a modular version of the app developed by Fernando in the ds4a (data science for all course) CorrelationOne project challenge. [https://gitlab.com/ferroao/nlpfeedback]
Credentials for mongodb are not available in this repository, following ds4a rules.
In the gitlab repository are the python notebook files (models), written by a group of ds4a students, including one author of this repository.
This repo includes a translation procedure for building a model for Africa community feedback (english)
The shiny app is connected to a source database of community feedback. This feedback is provided continuously to NGOs (into the database of a third party - here as PoC only)
The feedback should be constantly updated (form page), filtered (mod. filter), whenever needed tagged (tag module), based on a NLP model developed using a subset of tagged data (stored in .pkl files).
Finally, for the health and cash service types (models), text can be subset (mod. service_type).
This way NGOs can attend and prioritize solutions for communities.
There are models for the services: Healthcare (Africa and Colombia) and
Cash Transfer (Colombia). The python model is precomputed, saved in .pkl
files, see folder pkl, and loaded by python scripts in folder py,
from a call in the R module tag. The pertinent python notebook
developed in group is on the gitlab repository, see above. In the
notebook, the accepted model is SVC.
Dockerfilefilerequirements.txtfilepyfolder for scripts developed by Fernando- models are saved in
.pklfiles (pklfolder) - See the
tagmodule in foldermodulesfor the callssystem("python3...
- learn to use modules for better coding practices, performance
- learn rhino/box for better coding practices
- use shiny.fluent to learn to generate better UIs
- share in Shiny Conference as an example to community on those aspects