Content-based Influence Modeling for Opinion Behavior Prediction

Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pages 2207–2216, Osaka, Japan, December 11-17 2016. Content-based Influence Modeling for Opinion Behavior Prediction Chengyao Chen, Zhitao Wang, Yu Lei and Wenjie Li Department of Computing,The Hong Kong Polytechnic University {cscchen, csztwang, csylei, cswjli}@comp.polyu.edu.hk Abstract Nowadays, social media has become a popular platform for companies to understand their cus- tomers. It provides valuable opportunities to gain new insights into how a person’s opinion about a product is influenced by his friends. Though various approaches have been proposed to study the opinion formation problem, they all formulate opinions as the derived sentiment values either discrete or continuous without considering the semantic information. In this paper, we propose a Content-based Social Influence Model to study the implicit mechanism underlying the change of opinions. We then apply the learned model to predict users’ future opinions. The advantage of the proposed model is the ability to handle the semantic information and to learn two influence components including the opinion influence of the content information and the social relation factors. In the experiments conducted on Twitter datasets, our model significantly outperforms other popular opinion formation models. 1 Introduction Social media services, such as Twitter, Facebook, etc. provide fast and effective platforms for people to receive messages from their neighbors/friends and express their own opinions. The online commu- nication can gradually influence one’s opinions (Anagnostopoulos et al., 2008). In fact, according to a marketing survey1, 71% of the consumers said they are more likely to make a purchase based on social media referrals. Naturally, social media offers a great chance for companies to conduct marketing by in- fluencing the opinions of their potential customers. In order to achieve that, exploring and understanding the intrinsic mechanism of opinion formation is of great importance. Informational influence is a primary process for forming opinions on products in social media (Das et al., 2014). It describes the following scenario: when users lack the necessary information, they will seek for the opinions of their neighbors to update their beliefs. Taking the informational influence as premise, several models are proposed with different assumptions of how a person updates her/his own opinions according to the neighbors’ opinions (Clifford and Sudbury, 1973; DeGroot, 1974; Hegselmann and Krause, 2002). In these models, opinions are pre-defined as statuses through discrete categories including positive, negative and neutral opinion (Hegselmann and Krause, 2002; Galam, 2002; De et al., 2014) or continuous scales of opinion strengths (Clifford and Sudbury, 1973; DeGroot, 1974; Yildiz et al., 2011; Chazelle, 2012). However, on most social media platforms, people exchange their views by posting and replying through textual messages. The summarized opinion status simplifies the opinion formation process, and ignores the effects of semantic information hidden in the exchanged content information. Even if two messages have the same opinion category, different semantic information of the contents may result in different effects on others’ opinions. We take two postings of the product ”Samsung Galaxy” as examples: (1): ”I can’t post gifs …

Content truncated to 500 words