An Effective Way to Enhance Classifications for the Semi-Structured Research Articles

  • Ejaz Ahmed
  • Sumbal Ashraf
  • Waseem Shahzad
Keywords: Structured research articles, semi-structured research articles, supervised technique, Gensim, TFID


Due to the drastic increase in the research publications, numerous research articles are available electronically on different online digital libraries. Some research articles or papers are not retrieved during online searches due to their classification issues. The adequately structured research articles are relatively easily approachable as compared to semi-structured and unstructured research articles, and sometimes the reader does not get accurate results on different digital libraries as the research articles are not classified properly. Neglecting the semi-structured and unstructured published research not only causes gap deficiency but also affects the results of the proposed techniques and citations for other articles. Usually, researchers missed semi-structured and unstructured research articles during their online search. Classification techniques have been applied to structured articles and no significant work has been performed towards the classification of semi-structured and unstructured research articles. Therefore, this research focuses on the classification of semi-structured research articles using different supervised classification techniques so that the most accurate and large amount of relevant research results will be achieved. For experimentation, a labeled dataset was used for the classification of semi-structured papers. The dataset we used for experimentation is comprised of manually gathered research articles from Santos repository dataset and labeling them accordingly. The current study used four different supervised classification techniques such as Support Vector Machine (SVM) classifier, Naïve Based classifier, K Nearest Neighbor classifier, and Decision Tree classifier. The comparison was performed between these supervised classification techniques to see which classifier gives better accuracy. The unit of measures or parameters selected to compare these classifiers are: accuracy, recall, precision, and f-score. The evaluation was performed on the basis of results and comparison in the experimentation. Experimental results of classifiers K-neighbor are better than other classifiers SVM, Decision Trees and Naive Bayes

How to Cite
Ejaz Ahmed, Sumbal Ashraf, & Waseem Shahzad. (2020). An Effective Way to Enhance Classifications for the Semi-Structured Research Articles. University of Sindh Journal of Information and Communication Technology , 4(1), 17-23. Retrieved from