U3E: Unsupervised and Erasure-based Evidence Extraction for Machine Reading Comprehension

More tasks in Machine Reading Comprehension (MRC) require predicting the answer and extracting supported evidence sentences simultaneously. However, the annotation of supporting evidence sentences is usually time-consuming and labor-intensive. In this paper, to address this issue and consider that most of the existing extraction methods are semi-supervised, we propose an unsupervised and erasure-based evidence extraction method named U3E, which takes the changes after sentence-level feature erasure in the document as input, simulating the decline in problem-solving ability caused by human memory decline. In order to make selections based on fully understanding the semantics of the context, we also propose metrics to quickly select the optimal memory model for these input changes. To compare U3E with typical evidence extraction methods and investigate its effectiveness in evidence, we conduct experiments on different datasets. Experimental results show that U3E is simple but effective, also improves model performance.