Original sources of the initial NLI datasets

**SICK**
paper: https://aclanthology.org/L14-1314/
dataset: https://alt.qcri.org/semeval2014/task1/index.php?id=data-and-tools (the page lacks style files)
The dataset has three splits: train, trial, test
This differs from another [source](https://marcobaroni.org/composes/sick.html) as it contains slightly more problems and has been used by more systems.

**FraCaS**
paper: https://gu-clasp.github.io/multifracas/D16.pdf
dataset: https://nlp.stanford.edu/~wcmac/downloads/fracas.xml

**Stanford NLI (SNLI)**
paper: https://aclanthology.org/D15-1075/
dataset: https://nlp.stanford.edu/projects/snli/

While adding NLI problems from these datasets, the metadata of each problem needs to be determined, e.g., gold label, original id, annotator labels (if any), etc.
I am happy to work on it as I am well aware of the relevant meta data per dataset.