DeanLight
commented
3 years ago Current use cases (will update as more come in)
These are some ideas to play around with. We dont have to do all of them, i just put more than we need so we can throw out some if we see that they are too much work/ too similar in implementation to others.
Family tree
given a text about a family, generate their family tree.
- We can start with our own naive text
- We can later take texts from famous stories like a hundred years of solitude/ greek mythology and see how well we fared versus the existing family trees.
Medical compounds that help with covid/diease x
Given a couple of academic articles about coivd/disease x, find the all chemical compounds / drugs that were said to have a positive effect on patients.
animal classification
Given an encyclopedia about animals
- Generate the type system of the animal kingdom (dogs and cats are mammals etc ... )
- Generate the features set of each category This basically puts some set operation/ statistics on top of relations we extract. For example, we need to extract "is a" relations and "has a" (the shark has gills) relations and for each category show what features all or most of a group has.
nested text parsing of medical documents
Stjin gave us a use case of a real lab that needs to parse nested medical text from sensor software into a relational DB. I am attaching two examples of text and a pdf that outlines the schema of the DB 0.txt 1.txt dataset_descriptor.pdf
nnrepos
follows #43
brief:
we talked about our two main planned features - detailed tutorials and replacing pydatalog with sql. our current tasks are passing
feature-client-only's code review, and simultaneously planning/implementing tutorials.steps:
feature-client-onlybranch's issue with rule safetyfeature-client-onlyintomasterpytorch's tutorial as an examplethe next meeting is scheduled for may 10th