AI helps you reading Science

AI generates interpretation videos

AI extracts and analyses the key points of the paper to generate videos automatically

Go Generating

AI Traceability

AI parses the academic lineage of this thesis

Master Reading Tree
Generate MRT

AI Insight

AI extracts a summary of this paper

First we let our algorithm identify highly suspicious spammer candidates

Review Graph Based Online Store Review Spammer Detection

ICDM, pp.1242-1247, (2011)

Cited by: 374|Views184


Online reviews provide valuable information about products and services to consumers. However, spammers are joining the community trying to mislead readers by writing fake reviews. Previous attempts for spammer detection used reviewers' behaviors, text similarity, linguistics features and rating patterns. Those studies are able to identif...More



  • Online store reviews are an important resource to help people make wise choices for their purchases.
  • The authors have the following observations: 1) The authors can judge the honesty of a review given the reliability of the store it was posted to, plus the agreement of the review with its surrounding reviews about the same store.
  • A review on this store should have a high honesty score if it agrees with many other honest reviews.
  • Online store reviews are an important resource to help people make wise choices for their purchases
  • First we let our algorithm identify highly suspicious spammer candidates
  • We recruit human judges to make the judgments on the candidates about whether they seem to be real spammers
  • Similar evaluation approaches have been used in previous review spam detection research [3, 8]
  • We evaluate the top 100 suspicious reviewers identified by our model
  • Our method aims to find those that cannot be found by previous methods
  • The authors only consider reviewers with positive trustiness score because their ratings really reflect the store’s quality.
  • F. Iterative Computation Framework Integrating the pieces of information of the review graph together, the authors have an iterative computation framework to compute reliability, trustiness, and honesty, by exploring the inter-dependencies among them.
  • A spammer detection algorithm is effective, if different human evaluators agree with each other about their judgments and concur with the system on the same set of results.
  • 2) Human Evaluation Process: Judging suspicious spammers is a complex task for human and often involves intuition and searching for additional information, especially when the authors target at more subtle spamming activities.
  • To decide if a candidate is a spammer requires human judges to read his/her reviews and ratings, and to collect evidences from relations with other reviewers, stores, and even the Internet.
  • To standardize the judgment process, the human evaluators agree upon three conditions as the evidence to claim that a reviewer is a potential spammer.
  • A reviewer is suspicious if (s)he has a significant number of reviews saying opposite opinions about some stores as compared to evidences presented by general web search results.
  • It gathers reports on business reliability, alert the public to business or consumer scams, and enforce the mutual trustiness between consumers and companies resellerratings, business honesty information from BBB, and search results from Google, and by reading reviews.
  • Given the differences between the work and previous studies, the authors want to demonstrate that suspicious reviewers found by the method can hardly be identified by existing techniques.
  • They first look for candidates who have multiple reviews about one target, and compute spamming scores to capture spammers from the candidates.
  • This paper proposed a novel review graph model and an iterative method utilizing influences among reviewers, reviews, and stores to detect spammers.
  • The method showed how the information in the review graph indicates the causes for spamming and reveals important clues of different types of spammers.
  • Experimental results show that the proposed method can identify subtle spamming activities with good precision and human evaluator agreement.
Download tables as Excel
  • 1 This work is supported in part by NSF through grants IIS-0905215, OISE-0968341, DBI-0960443, and IIS-1111092
  • J.L. Fleiss and J. Cohen. The equivalence of weighted kappa and the intraclass correlation coefficient as measures of reliability. Educational and Psychological Measurement, 1973.
    Google ScholarLocate open access versionFindings
  • M. Hu and B. Liu. Mining and summarizing customer reviews. In KDD, 2004.
    Google ScholarLocate open access versionFindings
  • N. Jindal and B. Liu. Opinion spam and analysis. In WSDM ’08, [4] N. Jindal, B. Liu, and E.P. Lim. Finding unusual review patterns using unexpected rules. In CIKM, 2010
    Google ScholarLocate open access versionFindings
  • [5] J. Kleinberg. Authoritative sources in a hyperlinked environment. In Journal of the ACM (JACM), 1999.
    Google ScholarLocate open access versionFindings
  • [6] J.R. Landis and G.G. Koch. The measurement of observer agreement for categorical data. Biometrics, 1977.
    Google ScholarLocate open access versionFindings
  • [7] F. Li, M. Huang, Y. Yang, X. Zhu. Learning to identify review Spam. In IJCAI, 2011.
    Google ScholarLocate open access versionFindings
  • [8] E.P. Lim, V.A. Nguyen, N. Jindal, B. Liu, and H.W. Lauw. Detecting product review spammers using rating behaviors. In CIKM, 2010.
    Google ScholarLocate open access versionFindings
  • [9] Ott, M., Y. Choi, C. Cardie, and J.T. Hancock. Finding deceptive opinion spam by any stretch of the imagination. In ACL, 2011.
    Google ScholarLocate open access versionFindings
  • [10] X. Yin, J. Han, and P.S. Yu. Truth discovery with multiple conflicting information providers on the web. In TKDE, 2008. 4
    Locate open access versionFindings
Your rating :