Fig. 1From: Exploiting sequence labeling framework to extract document-level relations from biomedical textsAn annotated abstract in CDR corpus and instances constructed in classification-based methods. In the annotations, only related pairs are listed without specifying the exact entity offsets. There are 4 independent instances (entity pairs) constructed from the same sentence in the abstract according to classification-based methodsBack to article page