Comparison of Full-Text Indexing with Metadata Indexing Based Subject Classification Using Graph-Based Index

Abstract

Subject classification is an indispensable part of all academic search engines to facilitate faster search and retrieval of scholarly articles based on search queries. The widely used approach uses the metadata of journal papers like title, abstract, paper keywords, etc., to classify articles. This paper compares full text-based subject classification with metadata-based subject classification using a graph-based indexing approach. Comparing both methods is an extension of my previous work, GASE, a Graph-based Academic Search Engine based on the subject classification of research articles using an efficient full-text indexing approach. The results show that full-text indexing-based subject classification yields high accuracy than metadata-based classification. Also compared the space complexity and time complexity of both indexing methods. Full-text indexing will have higher space complexity, as it requires storing the entire contents. But subject labeling takes up a generalized time complexity of ? (n2 log(n) 2) for both full-text and metadata indexing by considering only the higher-order term and ignoring other constant values.