Discourse analysis based credibility checks to online reviews using deep learning based discourse markers. (March 2023)

• Record Type: Journal Article
• Authors: Alawadh, Husam M.; Alabrah, Amerah; Meraj, Talha; Rauf, Hafiz Tayyab
• Is Part Of: Computer speech & language. Volume 78(2023)
• Publication Date: 2023-03
• Subjects: Credibility check -- Discourse analysis -- Discourse markers -- Deep learning -- Machine learning; Speech processing systems -- Periodicals; Automatic speech recognition -- Periodicals; Computers -- Periodicals; Linguistics -- Periodicals; Speech-Language Pathology -- Periodicals; Traitement automatique de la parole -- Périodiques; Reconnaissance automatique de la parole -- Périodiques; Automatic speech recognition; Speech processing systems; Electronic journals; Periodicals; 006.454
• DOI: 10.1016/j.csl.2022.101450
• Languages: English
• ISSNs: 0885-2308

Abstract

Abstract: E-commerce strongly relies on marketing strategies that utilize user given feedback on services provided. Hotels and other hospitabilty outlets such as restaurants enlist user feedback about their services in their online reservation systems, thereby helping thier prospective customers in the decision making process. The credibility of such reviews is of paramount significance and any compromise of the process can lead to unwanted consequences both for customers and businesses. As such, the need for a method by which deceptive reviews are identified is evident to protect community and businesses alike. Deceptive reviews typically exhibit distinctive linguistic clues which can be detected, raising a flag about the credibilty of the user who posted them . Specifically, discursive features that show in such texts can be a useful tool in checking the credibility of reviews. Manually analysing the discourse and different rhetorical structures can be a tedious and timeconsuming process allowing some deceptive reviews to remain publically available. However, coupled with artificial intelligence tools such as deep learning approaches, discourse analysis could be performed in a manner that is both efficient and timely. The proposed study used a balanced publically available deceptive and truthful reviews dataset to design a discourse analysis based credibility check scheme with high accuracy. The multi-context of text is extracted via multi-n-gram windows (named as Discourse markers) applied in a proposed deep multi-channel convolutional neural network (MC-CNN). The hold-out approach is applied with 6 various splits of data on the proposed method MC-CNN to validate the performance. From various experiments, we report the best F1-score 87% has been achieved that not only defended the deep learning based discourse analysis but also defended the hypothesis made on deceptive and truthful reviews by proposed study. Highlights: The hypothetical differences on real and fake reviews are highlighted and discussed. Intelligent and automated solution via ML technique is proposed to check credibility. The enhanced and improved machine learning method for discourse analysis is proposed.