Fine-tuned T5 Models on FairytaleQA Chinese Dataset

Abstract

Question answering (QA) training plays a significant educational role in boosting comprehension learning skills for both machines and young children. One representative of QA datasets is FairytaleQA, but it is limited to English. Therefore, in this project, we manually translate the essence of the dataset into Chinese and validate the utility of the FairytaleQA Chinese dataset based on five T5 models. The results demonstrate that the FairytaleQA Chinese dataset has acceptable utility.