Integrating aspect analysis and local outlier factor for intelligent review spam detection

摘要

Recently, online reviews are used increasingly by individuals and organizations for making purchase and business decisions. Unfortunately, driven by profit and fame, spammers post spurious reviews to mislead the customers. Therefore, intelligent detection of the spam reviews from a large scale of texts is a great challenge. This paper describes an unsupervised method aiming for intelligently detecting online review spams. Review spam detection is transformed into a density-based outlier detection problem. The proposed method generates a sentiment lexicon to calculate the aspect rating of reviews,and proposes an aspect-rating local outlier factor model (AR-LOF) to identify the spam reviews. The experiments on TripAdvisor demonstrate the high effectiveness and intelligence of the proposed model, which has the potential to significantly help the online web business. (C) 2019 Elsevier B.V. All rights reserved.