A Structural SVM Approach for Reference Parsing

MEDLINE®, the flagship database of the U.S. National Library of Medicine, is a critical source of information for biomedical research and clinical medicine. The automated extraction of bibliographic data, such as article titles, author names, abstracts, and references, is essential to the affordable creation of this citation database. References, typically appearing at the end of journal articles, can provide valuable information for extracting Comment-On/Comment-In data (identifying commentary article pairs) and assigning MeSH terms in an article. The regular structure in references enables us to implement structural SVM, a newly developed structured learning algorithm to parse references. In this study we use two types of contextual features to compare structural SVM with conventional SVM. When only basic observation features are used for each token, structural SVM achieves higher performance compared to SVM since it utilizes the contextual label features. However, when the contextual observation features from neighboring tokens are combined, SVM performance improves greatly, and is close to that of structural SVM after adding the second order contextual observation features. Both methods achieve above 98% token classification accuracy and above 95% overall chunk-level accuracy for reference parsing.