Using Deep Linguistic Features for Finding Deceptive Opinion Spam

Abstract

While most recent work has focused on instances of opinion spam which are manually identifiable or deceptive opinion spam which are written by paid writers separately, in this work we study both of these interesting topics and propose an effective framework which has good performance on both datasets. Based on the golden-standard opinion spam dataset, we propose a novel model which integrates some deep linguistic features derived from a syntactic dependency parsing tree to discriminate deceptive opinions from normal ones. On a background of multiple language tasks, our model is evaluated on both English (gold-standard) and Chinese (non-gold) datasets. The experimental results show that our model produces state-of-the-art results on both of the topics. TITLE AND ABSTRACT IN ANOTHER LANGUAGE (MANDARIN) Т5-Ã&E ¥ p?ŠóA &¢