Using Incremental Addition to Evaluate the Grouping Quality of Document Classification

Most previous research on document classification assigned only one or two category tags to each document. Furthermore, the tagged items are rarely incorporated into their topic groups in subsequent classification work though it would conceivably enhance classification efficiency. with the modularity method, this research incrementally adds the classified documents to their topic groups after the recognition process to examine the changes in grouping quality. the result shows that social network analysis demonstrates great potential for automatic document classification, especially in identifying citation networks embedded in research papers and reference lists. a modified TF-IDF technique calculates the weight of each keyword in the topic groups. All the papers under study are collected from three journals in IEEE Computer Society collection published from 1979 to 2011.