Discovering the Latent Writing Style from Articles: A Contextualized Feature Extraction Approach

Computational Linguistics and Chinese Language Processing Vol. 24, No. 1, June 2019, pp. 15-38 15  The Association for Computational Linguistics and Chinese Language Processing Discovering the Latent Writing Style from Articles: A Contextualized Feature Extraction Approach Yen-Hao Huang, Ting-Wei Liu, Ssu-Rui Lee, Ya-Wen Yu, Wan-Hsuan Lee, Fernando Henrique Calderon Alvarado and Yi-Shin Chen Abstract With the growth of the Internet, the ready accessibility and generation of online information has created the issue of determining how accurate or truthful that information is. The rapid speed of information generation makes the manual filter approach impossible; hence, there is a desire for mechanisms to automatically recognize and filter unreliable data. This research aimed to create a method for distinguishing vendor-sponsored reviews from customer product reviews using real-world online forum datasets. However, the lack of labelled sponsored reviews makes end-to-end training difficult; many existing approaches rely on lexicon-based features that may be easily manipulated by replacing word usages. To avoid this word manipulation, we derived a graph-based method for extracting latent writing style patterns. Thus, this work proposes a Contextualized Affect Representation for Implicit Style Recognition framework, namely CARISR. Transfer learning architecture was also adapted to improve the model’s learning process with weakly labeled data. The proposed approach demonstrated the ability to recognize sponsored reviews through comprehensive experiments using the limited available data with 70% accuracy. Keywords: Reliability, Transfer Learning, Writing Style, Text Classification, Natural Language Processing.  Department of Computer Science, National Tsing Hua University E-mail: yenhao0218@gmail.com 16 Yen-Hao Huang et al. 1. Introduction With the popularization of the Internet and communication devices, information can be sent more quickly and widely than ever before. However, technological advances have also made it difficult to avoid incorrect information. Sponsored reviews, which have recently become a popular marketing strategy in online forums, can provide incorrect information. The intention of these articles is to give their consumers a positive impression of the product. Some advertisement companies have even begun to use sponsored reviews as a new method of promoting their commodities. Such sponsored reviews usually only provide positive information about a product. Thus, these reviews may hide the disadvantages of a product and potentially mislead consumers into making an unbeneficial purchase. As unreliable data may contain incomplete or incorrect information, it is important to avoid them. Most of the filtering approaches on online social platforms rely on mutual reviewing from users or human-designed rules. However, no matter which approach is used, automatic filtering is still limited due to the various methods of writing sponsored reviews and how quickly information is generated. Consequently, a system to automatically identify these kinds of information has become an important issue in the information reliability research field. In this work, we focus on recognizing the information reliability of review articles on online web platforms. Review articles are widely consumed by readers in order for them to purchase the best products. General filtering methods fail to address two main difficulties. First, current filters are easily fooled if the method only considers word-based …

Content truncated to 500 words