This repo is the official code for the paper: Read-All-in-Once (RAiO): Multi-layer Contextual Architecture for Long-Text Machine Reading Comprehension, which is accepted in Q1 journal: IEEE Access.
We inherit partly code from Hugging Face. It is a huge advantage if you are familiar with the Hugging Face code style. Here, We only publish entire code for NewsQA dataset. To apply our method for NLQuAD, there are some slight different. But it is very easy to modify.
-
Run file bash: run.sh.
-
Note: All of the parameters can be understood easily. The reader are suggested to read our paper thoroughly to obtain the best params, and work with Hugging Face code style in advance to run our code.
-
If you encounter the 'Out of memory problem, please try the GPU with a larger VRAM
If our code is helpful for you, please cite my paper: Read-All-in-Once (RAiO): Multi-layer Contextual Architecture for Long-Text Machine Reading Comprehension. The link for our paper: https://ieeexplore.ieee.org/abstract/document/10190566