Citation Context Analysis using Word-Graph

Word Graph Citation Analysis Method (WGCAM) is proposed to perform sub graph matching of citation sentences into important and not-important citation classes. Graph similarity measures used in this research are: Maximum Common Subgraph Node Similarity (MCSNS), Maximum Common Subgraph Directed Edges Similarity (MCSDES), and Maximum Common Subgraph Undirected Edges Similarity (MCSUES). Experiments were carried out on an annotated dataset of 465 citations using five distinct classifiers; Support Vector Machine SVM, Random Forest, Naive Bayes, K-Nearest Neighbors KNN and Decision Tree. Decision Tree classifier outperforms and achieves an accuracy of 0.99.