| Driven by the popularity of mobile phones, short messages (SMS) have become anintegral part of social communication, with an increasing number of fraudulent SMS diffusinginto our everyday life. Based on discourse information analysis, through both real-life andexperiment data, this thesis points out21effective cues to detect fraudulent SMS.This thesis has three research questions. The first is to classify the real-life SMS data intodifferent categories by cluster analysis, and then analyze their information structure andcontent. Furthermore, this thesis aims to find out what effective cues can detect fraudulentSMS from truthful SMS. Finally, this thesis intends to analyze the information strategiesemployed in the fraudulent SMS.This thesis combines qualitative and quantitative analysis together. On the one hand,based on the model of tree information structure by Professor Du and other linguistic cuesretrieved from previous experiment models employed by American scholar Zhou, this thesisanalyzes the fraudulent SMS data and provides a system of potential cues to detect them. Onthe other hand, this thesis not only conducts the experiment but also analyzes the real-lifeSMS data retrieved from CLIPS to compare with the experiment results. T-test of SPSS16.0isemployed to study the significance of the potential fraudulent SMS cues.It is found that DIA variables such as information knots and information levels bothdemonstrate high significance in detecting fraudulent SMS. Besides, quantity, complexity,non-immediacy and other linguistic cues are also effective despite the fact that they showdifferent levels of significance in different fraudulent SMS groups. As a result, deceivers doutilize language differently than truth tellers and that combinations of cues can improve theability to predict fraudulent SMS.This thesis is expected to protect people from fraudulent SMS and offer assistance forgovernment and mobile-phone operators to develop fraudulent SMS auto detectiontechnologies. |
PDF Full Text Request