Joker at SemEval-2018 Task 12: The Argument Reasoning Comprehension with Neural Attention

Proceedings of the 12th International Workshop on Semantic Evaluation (SemEval-2018), pages 1129–1132 New Orleans, Louisiana, June 5–6, 2018. ©2018 Association for Computational Linguistics Joker at SemEval-2018 Task 12: The Argument Reasoning Comprehension with Neural Attention 1Guobin Sui , 1Wenhan Chao , 2Zhunchen Luo∗ 1School of Computer Science and Engineering / Beihang University 2Information Research Center of Military Science / PLA Academy of Military Science 1Beijing 100191, China; 2Beijing 100142, China {suigb,chaowenhan}@buaa.edu.cn;zhunchenluo@gmail.com Abstract This paper describes a classification system that participated in the SemEval-2018 Task 12: The Argument Reasoning Comprehension Task. Briefly the task can be described as that a natural language “argument” is what we have, with reason, claim, and correct and incorrect warrants, and we need to choose the correct warrant. In order to make fully understand of the semantic information of the sentences, we proposed a neural network architecture with attention mechanism to achieve this goal. Besides we try to introduce keywords into the model to improve accuracy. Finally the proposed system achieved 5th place among 22 participating systems. 1 Introduction In recent years, as an extremely important part of argument mining, argument reasoning has received considerable research. The argumentation reasoning can be used in many situations such as automatic score, policy decision, stance detection, and many others (Habernal et al., 2018). The task can be described as follows in detail: Given an argument consisting of a claim and a reason, the goal is to select the correct warrant that explains reasoning of this particular argu- ment. There are only two warrants given and only one answer is correct. The correct warrant inferred from the argument has a supported relation with the argument. However the other warrant either opposes the argument or has no correlation with the argument. Actually, this task could be treated as a binary classification(A vs. B). The task could be regarded as an argumentative relation work. The argumentative relation mining aims at identifying relations of attack and support between natural language arguments in text, by clas- sifying pairs of pieces of text as attack, support or neither attack nor support relations (Cocarascu and Toni, 2017). The corrected warrant means supporting the argument, while the other means attacking the argument. So we refer to some literature methods on ∗ Corresponding author. the relationship between arguments. Cocarascu and Toni (2017) use Long Short Term Memory (LSTM) to classify the relations between arguments. Rockt¨aschel et al. (2016) propose neural network with attention mechanism, making neural networks interpretable. We infer to their methods and construct our model in new manner. In this paper, we use the LSTM networks with attention mechanism to construct the classification system. The following sections are arranged as follows. In section 2, we will give an overview on the task and have an analysis of the datasets. In section 3 we describe the system used in this paper and introduce some interesting attempt in detail. Section 4 introduces the experiment and results. Finally, we get conclusions and have an outlook of the future work. …

Content truncated to 500 words