Contextualized medication event extraction with striding NER and multi-turn QA

Tomoki Tsujimura, Koshi Yamada, Ryuki Ida, Makoto Miwa, Yutaka Sasaki

Abstract: Highlights•We propose a striding NER model to keep the document structure as much as possible.•We propose a QA-based model that uses the annotation guideline for human as input.•Our system got the highest end-to-end score at the n2c2 2022 Track 1 evaluation.