Add GIN paper

[JeanKaddour]

Thank you very much for creating this excellent repo! I wanted to suggest adding our recent work on vectorized treatments. Our estimator is based on a generalization of the Robinson decomposition and allows one to decompose causal effect estimation into separate, simpler optimization problems. We have not tested text treatments (yet), but the experimental results on small-world network/molecular graph treatments look promising. Hence, I think this could be highly relevant for NLP datasets too.

[rpryzant]

Thanks for your contribution! I also think this looks very promising for applicability to text. So far this repo is structured for people who are looking to learn about text + causal inference, so the papers should have existing text applications. Could you add a new section for research with promising potential connections to language?

[JeanKaddour]

Thank you very much for your response! I followed your suggestion and added a new section. Please let me know if this works.

[rpryzant]

Perfect -- thanks! Merged.