An Investigation into the Indexing Practices of Four Databases: Part One

After reading a paper by Gloria Leckie titled “Desperately Seeking Citations,” I assigned it three keyword terms I thought were representative of the subject matter. These terms were “information seeking behavior-undergraduates,” “scholarly research” and “academic libraries-role.” [1] After comparing these to four databases, LISTA, LISA, Library Lit and ERIC, results showed five very different systems of subject headings.

My own terms were most similar to those of LISA, but the simple fact that the article’s message is inexact and can be interpreted in multiple ways means that no two systems will ever be exactly the same as long as the terms are applied by human indexers. In addition, some of the differences can partially be explained by the subject coverage of the databases, while others are simply due to individual indexing behavior and preferences by the organizations. Overall, they correspond to varying levels of effectiveness for their user groups, and Library Lit and ERIC were observed to be the most useful.

First, the LISTA and LISA databases featured rather similar descriptor systems, and initially seemed to be more effective than Library Lit; however, upon further investigation, Library Lit’s mechanisms for user discovery by subject appear to be superior. LISTA’s keyword terms were “information retrieval, report writing and library orientation for college students,” while LISA featured the descriptors “libraries, undergraduates, information seeking behaviour, and Faculty.” The subject coverage of the three databases is quite similar, but LISTA’s focus on technology may be reflected in their choice of “information retrieval.” Library Lit, on the other hand, used the keyword terms “bibliographic instruction/college and university students” and “college and university libraries/relations with faculty and curriculum.”

I initially thought that these terms made Library Lit less effective because they were not likely to correspond to search terms users would actually use. For example, I found it much less likely for a user to come up with terms like “bibliographic instruction” or “libraries/relations with faculty” on their own in a subject search than “undergraduates” or “information retrieval” which I thought were quite plausible. After performing some test searches, however, I changed my opinion. Regarding the LISA descriptors, the article only comes up if “behaviour” is spelled in the British manner including the “u,” a search query which returns 1176 peer-reviewed articles. However, a search for “information seeking behavior” returns two article citations containing that descriptor. This suggests that LISA has problems with indexing inconsistencies.

Also, after a closer examination, LISTA’s keyword phrases do not appear to be very effective. Users are not likely to use the term “library orientation” to describe the research process described in the article, and it did not appear to me that “information retrieval” was as significant a part of its subject matter as “faculty” or “information seeking,” two terms which it did not include in their indexing. Also, the subject phrases included are very broad, leading the user to over 20,000 articles and there does not appear to be an effective way to narrow your results if you begin with a subject search. Because of this, the effectiveness of subject headings for resource discovery is low.

On the other hand, Library Lit’s descriptors, which initially appeared to correspond poorly with actual user queries, actually function quite effectively in the context of the content discovery keys they provide. For example, a search for “undergraduate students” a likely user query which does not appear as a descriptor in the Leckie article citation, brings up several relevant subject phrases in the left sidebar of the page. One of these is “Bibliographic instruction/College and university students” from earlier, which takes the user to a less intimidating 1301 peer-reviewed articles including the Leckie piece. Their system also allows for easy further narrowing.

Continue Reading this Paper

[1] For my keyword terms with dashes, I envisioned them as part of a hierarchy. In my hypothetical search system, the Leckie article would show up under a more general query for “information seeking behavior,” and users would see an option to further narrow the search.

Speak Your Mind