DiS-ReX: A Multilingual Dataset for Distantly Supervised Relation Extraction
Abhyuday Bhartiya, Kartikeya Badola, Mausam Mausam · 2022
Our goal is to study the novel task of distant supervision for multilingual relation extraction (Multi DS-RE).Research in Multi DS-RE has remained limited due to the absence of a reliable benchmarking dataset.The only available dataset for this task, RELX-Distant (Köksal and Özgür, 2020), displays several unrealistic characteristics, leading to a systematic overestimation of model performance.To alleviate these concerns, we release a new benchmark dataset for the task, named DiS-ReX.We also modify the widely-used bag attention models using an mBERT encoder and provide the first baseline results on the proposed task.We show that DiS-ReX serves as a more challenging dataset than RELX-Distant, leaving ample room for future research in this domain.