A hybrid approach to faceted classification based on analysis of descriptor suffixes

Abstract

This study explores the construction of a faceted vocabulary that can be used as a mechanism for organizing Web-based resources. After analyzing the manual process of faceted vocabulary construction using existing organizational structures to identify heuristics for automating the construction process, we modeled a hybrid, semi-automatic approach to facet generation that integrates the strengths of manual and automatic methods. This approach is based on the organization of terms according to the meanings of their suffixes.Although this heuristic could not usefully organize the majority of terms of the lexicon bases to which it was applied, it nevertheless shows promise as a component of hybrid classification.Enumerative classification schemes have long provided an effective tool for organizing a collection of resources by assigning each resource to a single class in a set of predefined and mutually exclusive classes. Although librarians have traditionally relied on classification schemes such as the Library of Congress Classification (LCC) and the Dewey Decimal Classification (DDC) to provide access to physical resources, the use of machine-based full-text searching undermined the perceived utility of classification for information discovery and retrieval. However, growing frustration with the huge retrieval sets and numerous false drops that accompanied do-it-yourself searching on the Web has generated renewed interest in classification, categorization and the power of controlled vocabularies --interest reflected across the Web landscape, from Web directories such as Yahoo! to metadata initiatives associated with digital libraries and the Semantic Web.There are many challenges to a classification-based approach to organizing the Web. For example, it is impossible to "organize" the whole Web due to its massive size and the diversity of Web resources. Even if such a feat were feasible, clustering approaches are not incremental and text categorization approaches are based on a static classification scheme, rendering them unable to deal with the dynamic nature of the Web corpus. A highly variable and dynamic environment such as the Web requires an organizational approach that not only provides flexibility of representation and accommodates the dynamic nature of human knowledge itself but is also able to respond to the information needs of a highly diverse and increasingly interdisciplinary population.Because traditional classification schemes attempt to enumerate all knowledge in a given domain within a fixed set of predetermined classes, they are ill-suited for organizing resources in the diverse and multidisciplinary environment of the Web. Recognizing the inherent rigidity of traditional enumerative structures, 1945) proposed a more flexible approach to organization that represented knowledge not as a set of static classes but as a set of concepts and relationships. This approach identifies the various aspects (characteristics or facets) of a given domain so as to derive a set of independent concept hierarchies t...

Combining Faceted Classification and Concept Search: A Pilot Study

“…Rather, analysis of domain content should be carried out within the existing conceptual framework of the domain, utilizing both inductive "bottom-up" and deductive "top-down" strategies (Loehrlein et al 2005;Yang and Jacob 2004). By beginning with a top-down investigation of the conceptual framework of the domain, subsequent analysis of domain terms and term relationships will be better able both to identify the most relevant concepts that will constitute the initial set of baseline facets and to establish the meaningful relationships that obtain between those facets.…”