Hey HN,
I'm a student at Harvard that is working in lab on a project in the area of disentangling representations, and I was wondering if there are open source tools that act as a sort of data layer in your pipeline, building useful downstream representations that your models can be trained on and used for a wider variety of tasks?
I know in the past tools like DeepDive from stanford worked on structuring unstructured data, but they really focused on making data tabular. Most data in the world can probably be split into more useful representations than that. Am I mistaken in any of my thoughts/assumptions?
