ECNU at SemEval-2018 Task 12: An End-to-End Attention-based Neural Network for the Argument Reasoning Comprehension Task

Proceedings of the 12th International Workshop on Semantic Evaluation (SemEval-2018), pages 1094–1098 New Orleans, Louisiana, June 5–6, 2018. ©2018 Association for Computational Linguistics ECNU at SemEval-2018 Task 12: An End-to-End Attention-based Neural Network for the Argument Reasoning Comprehension Task Junfeng Tian1, Man Lan1,2* and Yuanbin Wu1,2 1 School of Computer Science and Software Engineering, East China Normal University, Shanghai, P.R.China 2 Shanghai Key Laboratory of Multidimensional Information Processing 51151201048@stu.ecnu.edu.cn, {mlan, ybwu}@cs.ecnu.edu.cn Abstract This paper presents our submissions to Se- mEval 2018 Task 12: the Argument Reasoning Comprehension Task. We investigate an end- to-end attention-based neural network to rep- resent the two lexically close candidate war- rants. On the one hand, we extract their differ- ent parts as attention vectors to obtain distin- guishable representations. On the other hand, we use their surrounds (i.e., claim, reason, de- bate context) as another attention vectors to get contextual representations, which work as fi- nal clues to select the correct warrant. Our model achieves 60.4% accuracy and ranks 3rd among 22 participating systems. 1 Introduction Reasoning is a crucial part of natural language ar- gumentation. In order to comprehend an argumen- t, one must analyze its warrant, which explains why its claim follows form its premises (aka rea- sons) (Habernal et al., 2018a). SemEval-2018 Task 12 provides the argumen- t reasoning comprehension task (Habernal et al., 2018b). Given a reason and a claim along with the title and a short description of the debate they oc- cur in, the goal is to identify the correct warrant from two candidates. Figure 1 gives an example. The abstract structure of an argument is: reason → (since) warrant → (therefore) claim. The challenging factor is that both candidate warrants are plausible and lexically very close while leading to contradicting claims (Habernal et al., 2018a). Here we give three examples of the two candidate warrants: Ex1: A huge pandemic would (not) be a great news story. Ex2: The role of a citizen and a supreme court justice are inseparable /separable. Ex3: The rest of the comments can be skipped easily /make the section unreadable. Title: Have Comment Section Failed? Description: In recent years, many media companies have disabled them because of widespread abuse and obscenity. Reason: Many comments are thoughtful and insightful. And since {Warrant0 | Warrant1}, Claim: Comment sections have not failed.  Warrant0: The rest of the comments can be skipped eas- ily.  Warrant1: The rest of the comments make the section un- readable. Figure 1: An example of a debate in the argument rea- soning comprehension task. The differences are either negative words, or antonyms, or opposite phrases. Therefore, it is important to emphasize the different parts to ob- tain distinguish representations of the two war- rants, which express the opposite meanings. To address this factor, we proposed an end-to- end attention-based neural network. On the one hand, we extract the different parts of the two war- rants and use them as attention vectors to obtain warrants’ distinguishable representations. On the other hand, we …

Content truncated to 500 words