英文摘要 |
Along with the growing development of electronic information storage, textmining has increasingly gained attention from scholars and practitioners acrossvarious disciplines. In response to the need for meaning differentiation in socialstudies, the study aims to evaluate supervised machine learning classifiers interms of the performance of document classification. Setting out from the comparisonbetween traditional content analysis and text mining, the evaluation followsa normal procedure of text mining and applies Support Vector Machine andNaïve Bayes classifiers on non-structural, complex social texts extracted fromnews media. The outcomes of the analysis validate that text mining managesclassification well for documents with complex meaning. However, a further cowordnetwork analysis in the study finds that the editing style of data may affectclassifiers’ performance. It is suggested that, in the early stage of data processing,greater care must be given to the fit between research problems, editing styles,and classifiers. |