CDA-Based Spam Review Detection

With the development and popularization of the Internet,the promotion of national policies and the rapid development of the logistics industry,e-commerce has flourished in recent decades.The development of e-commerce has made commodity reviews more and more important,but spam reviews in commodity reviews interfere with people’s shopping experience.Therefore,it is necessary to effectively identify and eliminate spam reviews.Spam review on e-commerce platforms was the research object in this study,and the related research was conducted based on critical discourse analysis.In particular,this thesis addressed two research questions: 1)What are the stylistic features of spam reviews on e-commerce platforms? 2)Based on critical discourse analysis of spam review on e-commerce platforms,what are the reasons for the occurrence of spam reviews,and how is it related to power and ideology?Spam reviews collected in Taobao and published papers are the main data,and then the author examined them from two perspectives: stylistic features and critical discourse analysis.This thesis first analyzed stylistic features from three levels vocabulary,syntax and text,and then used Fairclough’s three-dimensional framework theory to analyze spam reviews from three aspects of description,interpretation and explanation.In terms of description,the transitivity and modality of spam review were analyzed.As for interpretation,the intertextuality of spam review was analyzed.In the perspective of explanation,social practice and discourse practice were analyzed,and the author attempted to explain the discourse power of spam review,the ideology behind it,and its relationship to society.Through the research,the author’s findings are as follows.Firstly,spam reviews on e-commerce platforms were low in directness and were not closely related to the product itself,as far as long spam reviews were concerned,the clauses in the texts were lack of coherence.Secondly,a phenomenon was emerging in spam reviews,in which merchants seized the voice of reviewers and turned them into tools for manipulating product reviews.Through the subsequent discoveries and research,there are some shortcomings in this paper.The research objective of this thesis is solely from e-commerce platforms in China.If other representative e-commerce platforms abroad can be used as the research object,the research can be more representative.The constructed corpus is small and corpus utilization is low.Therefore,further research is needed in the future.