Measuring the applicability of user-generated social tags along with expert-generated LCSH descriptors in Sociology: a heuristic study

Abstract

The study attempts to compare user-generated social tags with expert-generated LCSH descriptors of one thousand sociology books. The objective is to examine if social tags can be used to enhance the accessibility of library collections. The study found that both datasets do not follow the same vocabulary. Though, the Spearmans’ rank correlation (0.89) indicates a good association between common terms in both vocabularies. The Jaccard similarity coefficient (J = 0.13, 0.14, 0.17, 0.15 and 0.16) in different word clusters proves that top frequent social tags and top frequent LCSH descriptors used by users and experts are different. The comparison with each book also reveals that 555 books (55.5%) have 50 to 100 percent matching between both vocabularies. LCSH descriptor vocabulary contains more subject terms (24) than social tag vocabulary (12) out of the top thirty frequent terms. The comparison of social tags with MARC subfields ($a, $x, $y, $z, $v) reveals that users use more or less all the subfield terms as tags but either they do not use chronological terms ($y) for tags or use different terms other than experts for chronological information. Further, comparison with each book title reveals that social tags alongside LCSH descriptors can enhance the title-based search of libraries. Moreover, the study suggests that usage of social tags will not only enhance the accessibilities of library resources under sociology but also complement to controlled vocabularies by supplementing a variety of terms other than experts.