PICKLE Dataset

The PICKLE dataset accompanies the paper In a PICKLE: A gold standard entity and relation corpus for the molecular plant sciences. It is a natural language processing (NLP) dataset of scientific abstracts labeled with gold standard entities and relations. The abstracts were drawn from PubMed searches for the terms "jasmonic acid" and "gibberellic acid". There are 6,245 entities and 2,149 relations across the 250 documents in the brat-formatted (.txt/.ann) documents, and 6,164 entity and 2,094 relation annotations in the jsonl-formatted dataset, as some annotations cannot be aligned to the tokenization used in the jsonl format and are dropped.