Deceptive Review Spam Detection via Exploiting Task Relatedness and Unlabeled Data

Abstract

Existing work on detecting deceptive reviews primarily focuses on feature engineering and applies off-the-shelf supervised classification algorithms to the problem. Then, one real challenge would be to manually recognize plentiful ground truth spam review data for model building, which is rather difficult and often requires domain expertise in practice. In this paper, we propose to exploit the relatedness of multiple review spam detection tasks and readily available unlabeled data to address the scarcity of labeled opinion spam data. We first develop a multi-task learning method based on logistic regression (MTL-LR), which can boost the learning for a task by sharing the knowledge contained in the training signals of other related tasks. To leverage the unlabeled data, we introduce a graph Laplacian regularizer into each base model. We then propose a novel semi-supervised multitask learning method via Laplacian regularized logistic regression (SMTL-LLR) to further improve the review spam detection performance. We also develop a stochastic alternating method to cope with the optimization for SMTL-LLR. Experimental results on real-world review data demonstrate the benefit of SMTL-LLR over several well-established baseline methods.

A Review Structure Based Ensemble Model for Deceptive Review Spam

“…Donato et al [15] employed PU-learning to the problem using unlabeled data. Hai et al [16] developed a multi-task learning method based on logistic regression. Feng et al [17] studied the distributions of rating scores and introduced strategies to create a dataset with pseudo-standard.…”

Opinion Spam Detection based on Annotation Extension and Neural Networks

“…By comparison, Ott et al (2011) used AMT to crowdsource anonymous online turkers to construct the text-based spam review dataset; thus, they are real spams; several text-classification-based work have been performed based on this dataset (Feng et al 2012;Hai et al, 2016). Although expert annotation cannot fully eliminate the possibility of any noise, the spam reviews created by turkers may be more reliable because these reviews are real spams.…”

“…Although expert annotation cannot fully eliminate the possibility of any noise, the spam reviews created by turkers may be more reliable because these reviews are real spams. Due to the expensiveness of recruiting turkers, such kinds of existing datasets are usually limited (Hai et al, 2016). Li et al (2014) also released another spam review dataset, which has a style similar to that of Ott's dataset, but is still relatively small in scale.…”

Combining Deep Learning and Argumentative Reasoning for the Analysis of Social Media Textual Content Using Small Data Sets

“…Some previous work singles out quantity, specificity, diversity, non-immediacy, as well as task specific features such as affect, expressivity, complexity, uncertainty, and informality (Zhou et al 2004;Fuller et al 2006). Hai et al (2016) use review spam detection for different domains (hotel and restaurant) as a multitask learning problem by sharing the knowledge from training applied to each task and a graph regularizer for each model to incorporate unlabeled data. Mukherjee, Dutta, and Weikum (2017) use a model based on latent topic models in combination with limited metadata to compute a credibility score for reviews as well as to identify inconsistencies that appear between a review and the overall characterization of an item both for the item and for each latent facet.…”