Annotated dataset mentions corpus in IR/ML/NLP domain

This corpus is a re-annotated version of the dataset available at https://github.com/xjaeh/ner_dataset_recognition and described in the following publication:

Heddes, J.; Meerdink, P.; Pieters, M.; Marx, M. The Automatic Detection of Dataset Names in Scientific Articles. Data 2021, 6, 84. https://doi.org/10.3390/data6080084

The corpus contains 6000 sentences in the IR/ML/NLP domains, with dataset annotations.

The original corpus in CSV covered only explicitly named and reused datasets. In addition, "conjunctions" of datasets were annotated in a single span.

We review entirely the annotation to include new datasets too (as developed in the described research work of the source articles) and to annotate separately every individual datasets. In addition, we re-packaged the corpus into a more standard JSON with annotation offsets. Python scripts for conversion are available at https://github.com/kermitt2/dataset_recognition_resources

We thank the original authors of the corpus for their very valuable resource !