Looking Beyond Sentence-Level Natural Language Inference for Downstream Tasks
Anshuman Mishra Equal contribution. College of Information and Computer Sciences, University of Massachusetts Amherst Dhruvesh Patel* College of Information and Computer Sciences, University of Massachusetts Amherst Aparna Vijayakumar* College of Information and Computer Sciences, University of Massachusetts Amherst Xiang Li College of Information and Computer Sciences, University of Massachusetts Amherst Pavan Kapanipathi IBM Research Kartik Talamadupula IBM Research
Abstract
In recent years, the Natural Language Inference (NLI) task has garnered significant attention, with new datasets and models achieving near human-level performance on it. However, the full promise of NLI – particularly that it learns knowledge that should be generalizable to other downstream NLP tasks – has not been realized. In this paper, we study this unfulfilled promise from the lens of two downstream tasks: question answering (QA), and text summarization. We conjecture that a key difference between the NLI datasets and these downstream tasks concerns the length of the premise; and that creating new long premise NLI datasets out of existing QA datasets is a promising avenue for training a truly generalizable NLI model. We validate our conjecture by showing competitive results on the task of QA and obtaining the best reported results on the task of Checking Factual Correctness of Summaries.
中文速览
自然语言推理（Natural Language Inference, NLI）模型在大规模数据集上已接近人类水平，但把这些模型直接用到问答（QA）和文本摘要等下游任务时效果却很差。研究者发现症结在于：现有NLI数据集的前提（premise）几乎都是单句，而下游任务需要跨多句推理，两者所需的语言理解能力根本不在同一层次。为此，他们提出将现有多项选择阅读理解数据集通过规则、神经网络和混合三种方式转化为"长前提NLI数据集"，再用这些数据训练一个可迁移的NLI模型。实验表明，在多项选择阅读理解任务上，该方法达到了与专用模型相当的竞争力，在摘要事实性验证（Checking Factual Correctness of Summaries, CFCS）任务上更刷新了当时的最优成绩。这项工作揭示了前提长度才是制约NLI泛化能力的关键因素，为构建真正通用的NLI模型指出了一条可行路径。
原文 arXiv:2009.09099;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2009.09099v1