arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2208.14641cs.CLcs.AI
收藏

使用下一步监督生成NLI的中间步骤

Generating Intermediate Steps for NLI with Next-Step Supervision

Singapore University of Technology and Design(新加坡科技设计大学);IIT Kharagpur(印度理工学院卡拉格普尔分校);Microsoft(微软)

Deepanway Ghosal, Somak Aditya, Monojit Choudhury

摘要:

自然语言推理(NLI)任务通常需要通过多个步骤进行推理才能得出结论。虽然生成这些中间步骤(而不是摘要解释)的必要性已获得普遍支持,但不清楚如何在没有完整的端到端监督的情况下生成这些步骤,以及如何进一步利用这些生成的步骤。在这项工作中,我们训练一个序列到序列模型,在给定NLI前提和假设对(以及之前的步骤)的情况下,只生成下一个步骤;然后用外部知识和符号搜索来增强它,生成中间步骤,只有下一个步骤的监督。我们通过自动化和人工验证来证明这些生成步骤的正确性。此外,我们表明,这种生成的步骤可以帮助提高端到端的NLI任务的性能,使用简单的数据增强策略,在多个公共NLI数据集。

英文摘要:

The Natural Language Inference (NLI) task often requires reasoning over multiple steps to reach the conclusion. While the necessity of generating such intermediate steps (instead of a summary explanation) has gained popular support, it is unclear how to generate such steps without complete end-to-end supervision and how such generated steps can be further utilized. In this work, we train a sequence-to-sequence model to generate only the next step given an NLI premise and hypothesis pair (and previous steps); then enhance it with external knowledge and symbolic search to generate intermediate steps with only next-step supervision. We show the correctness of such generated steps through automated and human verification. Furthermore, we show that such generated steps can help improve end-to-end NLI task performance using simple data augmentation strategies, across multiple public NLI datasets.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑