Construction Of Feature Terms For Corruption Behaviours In Construction Projects

Corruption becomes seriously in the construction sector in China. Corruption isundetectable because of the projects’ unique complicity, the difficulty in determiningthe quality, plus the multiple participators and the time-consuming and complicateprocess as well as existing legislate loops. Governments, researchers, NGOs and otherstakeholders have been involved in anti-corruption cooperation.The current studies involve many defects including unclear distinction oncorruption in construction projects, segmentation in various disciplines and disperseddomain knowledge. Considering the useful information about corruptions inconstruction projects mainly exists in the form of unstructured text data, thetechnology of text mining would be a hotspot.This study is based on the National Natural Science Foundation of China“Research on Integrity Management Capacity of Project Execution Team”. Fixing onthe corruption behaviors in construction projects and citing the conception of domainknowledge, the thesis constructs the feature terms of corruption behaviors, includingmainbody, procesure and format, aimed at providing background knowledge for textmining in relative cases.This thesis clarifies corruption behaviors in construction projects based onentrusted power, and provides baseline for identifying corruption behaviors. Usingontology, social network analysis and semantic analysis, the article extracts separatelythe corruption mainbody, procedure and format from the core concepts of constructionstakeholders, basic procedures and formats. Thus, a three-dimensional describingmodel is constructed.Then, the model is used as the frame system of the feature terms. The thesiscombines two methods: the manually label relying on Grounded Theory and theautomatically key words extraction methods in order to extract the feature words ofmain body, procedure and format from the case texts describing corruptions.Furthermore, by contrasting the feature words extracted using these two methods,feature terms of the corruption behaviors in construction projects is built. The work isfurther completed by analyzing the relations of feature words, establishing verbal rules. Finally, the feature terms are used to cluster and classify the texts involving thecorruption behaviors and the results have shown better than the text mining methodbased on document only, thus proves the effectiveness of the feature terms.The research is basic work to develop the intelligent analysis on constructioncorruption.The achievement can be applied to the text mining based on semantics anddeal with corruption cases from vary approach. Besides, the domain knowledgeincluded in feature terms can also assist other text mining research in this field anddeepen our understanding of the corruption principles in construction sector.