Incorporating Argument-Level Interactions for Persuasion Comments Evaluation using Co-attention Model

Proceedings of the 27th International Conference on Computational Linguistics, pages 3703–3714 Santa Fe, New Mexico, USA, August 20-26, 2018. 3703 Incorporating Argument-Level Interactions for Persuasion Comments Evaluation using Co-attention Model Lu Ji1 , Zhongyu Wei2∗, Xiangkun Hu1, Yang Liu3 , Qi Zhang1 , Xuanjing Huang1 1School of Computer Science, Fudan University, Shanghai, China 2School of Data Science, Fudan University, China 3Liulishuo Company {17210240034,zywei,xkhu17,qi zhang,xjhuang}@fudan.edu.cn yang.liu@liulishuo.com Abstract In this paper, we investigate the issue of persuasiveness evaluation for argumentative comments. Most of the existing research explores different text features of reply comments on word level and ignores interactions between participants. In general, viewpoints are usually expressed by multiple arguments and exchanged on argument level. To better model the process of dialogical argumentation, we propose a novel co-attention mechanism based neural network to capture the interactions between participants on argument level. Experimental results on a publicly available dataset show that the proposed model significantly outperforms some state-of-the-art methods for persuasiveness evaluation. Further analysis reveals that attention weights computed in our model are able to extract interactive argument pairs from the original post and the reply. 1 Introduction Computational argumentation is a growing field in natural language processing. Existing research covers argument unit detection (Al-Khatib et al., 2016), argument structure prediction (Peldszus and Stede, 2015; Stab and Gurevych, 2014), argumentation scheme classification (Feng et al., 2014), etc. Recently, the automatic assessment of argumentation quality has started gaining attention. It can be analyzed at two levels, namely monological argumentation and dialogical argumentation. Monological argumentation refers to a composition of arguments on a certain issue (Wachsmuth et al., 2017). A typical example of quality evaluation for monological argumentation is automated essay scoring, which aims to process argumentative essays without human interference (Taghipour and Ng, 2016). It takes an essay as the input and outputs a numeric score, considering features of content, grammar, discourse structure and lexical richness (Burstein et al., 2013). Most of the efforts are made on the exploration for better document representation. Dialogical argumentation refers to a series of interactive arguments related to a given topic, involving argument retraction, view exchange, and so on (Besnard et al., 2014). With the popularity of online de- bating forums like convinceme 1, debatepedia 2 and change my view (CMV) 3, researchers pay increasing attention to evaluating the quality of debating or persuasive content (Tan et al., 2016; Wei and Liu, 2016; Wei et al., 2016; Habernal and Gurevych, 2016a). Although some similarity features of content between the reply and the original post are used to evaluate the quality of the given reply, they are computed on word level without considering the exchange of opinions on the basis of arguments. An example of dialogical argumentation is shown in Figure 1. There are three posts, one is original and the other two are replies. The repliers post to change the view of the original poster. And the positive reply is deemed to be more persuasive than the negative reply. We have three observations. First, viewpoints of the original …

Content truncated to 500 words