Identifying Comparative Claim Sentences in Full-Text Scientific Articles

Comparisons play a critical role in scientific communication by allowing an author to situate their work in the context of earlier research problems, experimental approaches, and results. Our goal is to identify comparison claims automatically from full-text scientific articles. In this paper, we introduce a set of semantic and syntactic features that characterize a sentence and then demonstrate how those features can be used in three different classifiers: Naive Bayes (NB), a Support Vector Machine (SVM) and a Bayesian network (BN). Experiments were conducted on 122 full-text toxicology articles containing 14,157 sentences, of which 1,735 (12.25%) were comparisons. Experiments show an F1 score of 0.71, 0.69, and 0.74 on the development set and 0.76, 0.65, and 0.74 on a validation set for the NB, SVM and BN, respectively.