DNC: probling NLI models with recasted datasets

My first NLP project. We recasted a lot of NLP datasets into the NLI format - premise and hypothesis, and used them to probe what kind of inferences NLI models are good at and what aren’t. I worked on recasting the VerbNet dataset. A fun project, and I learned a ton about NLP.

The paper is titled “Collecting Diverse Natural Language Inference Problems for Sentence Representation Evaluation,” which is presented at EMNLP 2018 by Adam.

Check it out on arXiv: https://arxiv.org/abs/1804.08207