Katerina Katsarou, Sukanya Sunder, et al.
SNAMS 2021
Multi-hop reading comprehension focuses on one type of factoid question, where a system needs to properly integrate multiple pieces of evidence to correctly answer a question. Previous work approximates global evidence with local coreference information, encoding coreference chains with DAG-styled GRU layers within a gated-attention reader. However, coreference is limited in providing information for rich inference. We introduce a new method for better connecting global evidence, which forms more complex graphs compared to DAGs. To perform evidence integration on our graphs, we investigate two recent graph neural networks, namely graph convolutional network (GCN) and graph recurrent network (GRN). Experiments on two standard datasets show that richer global information leads to better answers. Our approach shows highly competitive performances on these datasets without deep language models (such as ELMo).
Katerina Katsarou, Sukanya Sunder, et al.
SNAMS 2021
Paula Harder, Venkatesh Ramesh, et al.
EGU 2023
Maeda Hanafi, Yannis Katsis, et al.
EMNLP 2022
Richard Chen, Rameswar Panda, et al.
ICLR 2022