Logic-Driven Context Extension and Data Augmentation for Logical Reasoning of Text

Findings of the Association for Computational Linguistics: ACL 2022, pages 1619 - 1629 May 22-27, 2022 c⃝2022 Association for Computational Linguistics Logic-Driven Context Extension and Data Augmentation for Logical Reasoning of Text Siyuan Wang1∗, Wanjun Zhong2, Duyu Tang3, Zhongyu Wei1,5, Zhihao Fan1, Daxin Jiang3, Ming Zhou4, Nan Duan3 1School of Data Science, Fudan University, China 2Sun Yat-Sen University, China 3Microsoft, China 4Sinovation Ventures, China 5Research Institute of Intelligent and Complex Systems, Fudan University, China {wangsy18,zywei,fanzh18}@fudan.edu.cn; zhongwj25@mail2.sysu.edu.cn zhouming@chuangxin.com; {dutang,djiang,nanduan}@microsoft.com Abstract Logical reasoning of text requires identifying critical logical structures in the text and per- forming inference over them. Existing methods for logical reasoning mainly focus on contex- tual semantics of text while struggling to explic- itly model the logical inference process. In this paper, we not only put forward a logic-driven context extension framework but also propose a logic-driven data augmentation algorithm. The former follows a three-step reasoning paradigm, and each step is respectively to extract logical expressions as elementary reasoning units, sym- bolically infer the implicit expressions follow- ing equivalence laws and extend the context to validate the options. The latter augments lit- erally similar but logically different instances and incorporates contrastive learning to better capture logical information, especially logical negative and conditional relationships. We con- duct experiments on two benchmark datasets, ReClor and LogiQA. The results show that our method achieves state-of-the-art performance on both datasets, and even surpasses human performance on the ReClor dataset. 1 1 Introduction Recent years have witnessed a growing interest in logical reasoning of text, which learns to un- derstand a given text in logical level and perform logical inference to deduce implications from as- serted ones (McCarthy, 1989; Nilsson, 1991). As a significant component of human reading compre- hension, it is essential in many application scenar- ios, such as negotiation and debate. And several datasets have been proposed as benchmarks for this task (Williams et al., 2017; Habernal et al., 2017; Yu et al., 2020; Liu et al., 2020). An example of logical reasoning problems is shown in Figure 1, which takes a context descrip- ∗Work is done during internship at Microsoft. Zhongyu Wei and Duyu Tang are corresponding authors. 1Codes are publicly available at https://github. com/WangsyGit/LReasoner. Logical Symbols : 𝜶 : have keyboarding skills 𝜷 : be able to use a compute 𝜸 : be able to write your essays using a word processing program Extend the Implicit Logical Expressions by Laws: Context: If you have no keyboarding skills at all, you will not be able to use a computer. And if you are not able to use a computer, you will not be able to write your essays using a word processing program. Question： If above statements are true, which one of the following must be true? Options： A. If you are not able to write your essays using a word processing program, you have no keyboarding skills. B. If you are able to write your essays using a word processing program, you have at least some keyboarding skills. C. If …

Content truncated to 500 words