This code was used to create the dataset in: "Annotating Relation Inference in Context via Question Answering" Omer Levy and Ido Dagan. ACL 2016. Here is a brief description of each directory: assertions - Downloads Google's syntactic ngrams and extracts assertions from it. Run get_props.sh. qa - Given questions and a set of assertions, creates question-answer pairs for annotation. website - The user-interface used to annotate the task. lemma_baseline - The simplest baseline used in our evaluation.