Improving Intellectual Access to Material: An Online Browser for the Dewey Decimal Classification System.

Abstract

A report of research aimed at developing a technique that will automatically group similar books into clusters by subject and then organize these clusters hierarchically using the Dewey Decimal Classification Scheme so that users can broaden or narrow their searches at will. The software we have developed, uses an object oriented graphical user interface to facilitate browsing though the schedules. We studied the relationship between class numbers and subject headings and determined that the class number addresses the major focus of the work while the first subject heading represents the aboutness of the book. They are different and complementary. It was also evident that there was a classic many to many relationship between Dewey's and subject headings. We therefore decided to develop clusters based on the unique class number/subject heading combinations. We will discuss our experience with determining the optimum cluster size in a database of 100,000 records. Having manipulated the cluster size by altering the number of digits in the Dewey number, we then looked at the subject headings. Each subject heading has the potential for one or more subdivisions. We found examples with as many as 4 subdivisions or facets as we have chosen to call them. Out of the 65,904 subject headings only 16,290 had no facets. This means that 75% of the headings have one or more facets, while 6,291 or 14% of the headings had two or more facets. Our research has focused on search systems that manipulate subject clusters by combining subject headings and class numbers in new and interesting ways. Particularly powerful is the ability to keyword search through Dewey to find the relevant class number and then to be able to display the subject headings of books in that class. More interesting is the study of the clustering behavior in the system.