Synthetic Data Augmentation for Zero-Shot Cross-Lingual Question Answering
Arij Riabi‡ Thomas Scialom⋆⋄∗ Rachel Keraron⋆ Benoît Sagot‡ Djamé Seddah‡ Jacopo Staiano⋆ ‡ Inria, Paris, France ⋄ Sorbonne Université, CNRS, LIP6, F-75005 Paris, France ⋆ reciTAL, Paris, France ∗: equal contribution. The work of Arij Riabi was partly carried out while she was working at reciTAL.
Abstract
Coupled with the availability of large scale datasets, deep learning architectures have enabled rapid progress on Question Answering tasks. However, most of those datasets are in English, and the performances of state-of-the-art multilingual models are significantly lower when evaluated on non-English data. Due to high data collection costs, it is not realistic to obtain annotated data for each language one desires to support.
中文速览
跨语言问答系统长期面临一个核心困境：高质量的标注数据几乎只存在于英语，导致多语言模型在非英语语言上的表现大幅落后。研究者提出利用"问题生成"（Question Generation）模型来自动合成跨语言训练数据——先用英文SQuAD数据训练一个问题生成器，再结合机器翻译让它在多种语言的维基百科文本上批量生成"合成问答对"，随后用这些合成数据对问答模型进行预训练，再在英文数据上精调。实验表明，该方法在MLQA、XQuAD、SQuAD-it（意大利语）和PIAF（法语）四个多语言基准上均超越了此前最优结果，甚至在训练阶段从未见过的语言（如韩语）上也取得了显著提升。这项工作的价值在于，它无需额外人工标注就能有效提升低资源语言的问答能力，为多语言乃至低资源语言的问答系统开发提供了一条切实可行的数据增强路径。
原文 arXiv:2010.12643;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2010.12643v2