 |
Home | Search | Browse | About IPO | Staff | Links |
|
Home | Search | Browse | About IPO | Staff | Links |
|
Evaluating Selected Full-text Databases for Collection Development Introduction Like many libraries in the 1990s, Booth Library has been rapidly acquiring access to full-text databases. Through the generosity of the Illinois State Library, we have access to several wonderful full-text databases: ABI/Inform, EBSCO, FastDoc, Periodical Abstracts and WilsonSelect. Consortial arrangements have made it possible for us to subscribe to IDEAL, Project Muse and two IAC Info Trac databases—Expanded Academic Index Searchbank and Business and Company File. But what exactly do we have? While anecdotal information is plentiful, vendor descriptions lack specificity, and we are left with questions, including: What full-text titles are included in each database? How extensive is full-text title overlap between databases? What are the full-text subject strengths of each database? and What are the implications for collection development? This article results from a program presented at the Illinois Association of College and Research Libraries' 1998 Spring Conference. The program entitled "What's Hot, What's Not: Evaluating Electronic Journal Collections," was based on a study conducted during the fall semester of 1997 at Eastern Illinois University's Booth Library. This study examined nine full-text databases to determine the content and subject coverage of each database. We hoped that determining exactly what we had would lead us to a greater understanding of collection development possibilities in this new environment. Methodology In the data-gathering phase of the study, full-text title lists for each database were extracted from various vendor Web sites and organized in a database program. We then recorded a Library of Congress classification number for each full-text journal title. Numbers were taken from ILLINET Online whenever possible. If no LC call number was found in ILLINET Online, we consulted WorldCat and Ulrich's International Periodical Directory. If the class number was still unknown, an LC class number was assigned based on the journal's main subject coverage. Numbers were tallied by broad LC classification to build a subject profile for each database. See Table 1 for a list of LC classifications and subjects. Title overlap was determined by matching ISSN's. Similar methodology was used in a title coverage study conducted at John Carroll University's Grasselli Library.1 The purpose of this study was to examine title overlap of seven indexes to religious periodicals, using lists of periodicals indexed for 1994. Like our study, one of the goals of the investigation was to provide practical guidance for collection development. Database Profiles What did the profiles of each database reveal? We looked at Project Muse (PM) and IDEAL (ID) first. Both are 100 percent full-text and publisher-produced databases. The titles in a publisher-produced database tend to be more stable because the database producer controls the content. The online version of the journal is virtually identical to the print version of the journal, including accurate pagination and all graphics, sometimes in full color like Project Muse. However, the online versions usually do not include advertisements or other "extraneous" materials. IDEAL contained 177 journal titles produced by Academic Press. IDEAL (Figure 1) was extremely strong in the sciences (Q), had good content in bio-medical research (R), technology (T) and some business and psychology (H-HJ, B). By comparison, Project Muse was quite small, 43 titles produced by
* Karen Whisler, Head of Collection Management Services; Marlene Slough, Head of Circulation Services; Nackil Sung, Head of Library Technology Services; and Barbara Cressman, Assistant Head of Circulation Services at Booth Library, Eastern Illinois University, Charleston. 239 Johns Hopkins University Press. Project Muse (Figure 2) was strong in the humanities, particularly language and literature (P), but it also included titles in other subject areas, for example, the American Journal of Mathematics (Q). Seven of the databases studied, ABI/Inform (AB), Business & Company File (BC), EBSCO (EB), FastDoc (FD), Periodical Abstracts (PA), Expanded Academic Index Searchbank (SB) and WilsonSelect (WS), are generally referred to as aggregate databases. Unlike Project Muse and IDEAL, which have unique title content, these databases pull together articles from many sources. The database producer typically does not produce the database content, but has a contract with the journal publisher that allows the database producer to supply the publisher's content to us, the database purchasers. Publishers may sell their content to many database producers, leading to title overlap. Figure 3 shows the extent to which full-text titles were duplicated in the aggregate databases we studied. In comparing individual databases, overlap varied from about 10 percent to more than 90 percent. The aggregate databases exist in a very fluid environment. Titles are constantly being added or dropped, and their full-text status may change. Despite the uncertainties this caused, we did a subject analysis on each database using the full-text title content at the time of the study. What did we find in our study of full-text titles in the aggregate databases? ABI/Inform (Figure 4) and Business and Company File (Figure 5) were both strong in business (H-HJ). Another strength was found in technology, with both having more than 100 titles in class T. They were similar in size: ABI had 642 titles, Business & Company had 571 titles. While we expected the databases to be very similar, as illustrated in Figure 3, they had only 40 percent overlap in their full-text titles. Looking at individual titles, Business and Company had more regional business journals— Business and Company had 60 titles while ABI had only 40. Regional business titles are not typically in our print collection because they are hard to obtain and we tend to collect only for our own region, but they are very useful for students looking for business trends, case studies, etc. ABI was stronger in business law and health care administration. The subject coverage of EBSCO (Figure 6) and FastDoc (Figure 7) was virtually identical, both having 92 percent of their full-text titles in common. Subject coverage was very broad, with titles representing every LC classification. There were subject strengths in business (H-HJ), medicine (R), social sciences (HM-HX), technology (T). Of the seven aggregate databases, these two were the strongest in general works (A) and philosophy, psychology, and religion (B). They differed in that EBSCO indexed more than 4,000 titles, of which 27 percent were full-text, while FastDoc indexed 1,069 titles and was 100 percent full-text. Even though EBSCO offered more comprehensive indexing, students seemed to prefer the instant gratification offered by the complete full-text retrieval in FastDoc. Expanded Academic Index (Figure 8) and Peoriodical Abstracts (Figure 9) also were very similar, with nearly 50 percent full-text title overlap. They were close in size; Periodical Abstracts had 703 titles with a few more titles in education (L) and medicine (R) but each had titles in nearly every classification. Expanded Academic Index was the only exception, having no titles in naval science (V). Expanded Academic Index had 629 full-text titles. From our observations while working at the reference desk, both databases were useful for undergraduate research. As an Info Trac product, Expanded Academic Index enjoyed greater name recognition with students who may have encountered Info Trac in high school or junior college. They were both about one-third full-text, two-thirds citations. Of the seven aggregate databases studied, WilsonSelect seemed truly unique. WilsonSelect was most similar to Expanded Academic Index in title coverage, with 40 percent overlap in their full-text titles. With 400 titles, 100 percent full-text, drawn from popular journals indexed in H. W. Wilson indexes like Readers' Guide, Social Sciences Index, Humanities Index, Business and Periodicals Index, General Science Index, Education Index, etc., WilsonSelect appeared to be well suited to undergraduate research needs. It had strengths in business (H-HJ) and technology (T), but really had something for everyone (Figure 10). The content focused on popular topics, and the size insured adequate but not overwhelming retrieval sets. One hundred percent full-text meant no waiting! Students loved it! Finally, we also looked closely at title overlap and subject strengths comparing the electronic megacollection, that is viewing all nine databases collectively, to our print collection (Figure 11). In the megacollection, the overwhelming subject strength was in business (H-HJ), but there were also strengths in science, medicine 240 and technology (Q, R, T). In overall numbers, the megacollection was larger, with 2,772 full-text titles compared to our 2,507 active print titles. In subject coverage, the megacollection had more titles in general works (A), substantially more in business (H-HJ), and a few more in social sciences (HM-HX), medicine (R) and technology (T). Our print collection had more titles in history (E-F), geography, anthropology, recreation (G), education (L), music (M), art (N), literature (P), science (Q) and bibliography (Z). Collection Development What are some of the implications for collection development? We calculated our current subscription costs for the 746 full-text titles duplicated in our active print collection. Our current subscription costs for the duplicated titles in paper and/or microfilm totaled more than $73,000. This is worth repeating: more than $73,000! At Eastern, individual subject bibliographers manage the serials in their subject areas. Using the data collected in this study, reports were generated for each bibliographer showing the journal title, subscription costs and the databases where the title appears in full-text, with start dates if available. The reports may assist subject bibliographers as they make decisions about the cancellation of duplicate titles and the acquisition of titles not currently in our collection. To help increase the use of the full-text titles available, a list of titles linked to databases was generated for interlibrary loan, periodicals and reference. We also are investigating methods to add the full-text titles to our periodicals printout, a list of our periodical holdings generated by the Acquisitions Department. The list is arranged alphabetically by journal title and gives indexing information, summary holdings, call numbers and location. We hope to add the full-text titles to the periodical list with an abbreviated description of where full-text for individual titles might be found. However, a stumbling block in preparing the periodical list is trying to find appropriate wording to accurately reflect the nature of virtual subscriptions. It is important to keep in mind the problems common to the aggregate full-text databases: instability of full-text coverage, missing tables, incomplete pagination, etc. All these problems and more have been fully explored in articles like Orenstein's "The Fullness of Full-Text,"2 Ogburn's "The Full-Text Question,"3 and Grzeszkiewicz and Hawbaker's "Investigating a Full-text Journal Database." 4 We must make patrons aware of these virtual subscriptions, but we don't want to mislead students or promise more than we can deliver. Even in our print collection, we cannot absolutely promise that any given article will be available when needed. The article could be checked out, in use in the building, gone to binding, never received or ripped out. In the aggregate databases the article also could be missing for a variety of reasons. Our Cataloging Department is planning to catalog the titles in the more stable publisher-produced databases, but at this time we do not feel comfortable cataloging individual titles in the aggregate databases. Evaluating and managing full-text databases is an issue that cuts across all types of libraries of all sizes. Of paramount concern is how the evaluation of electronic resources affects the evaluation of the entire collection. The literature reveals a split between preference to use existing collection development policies for electronic resources or to implement a policy that would direct decisions about choosing formats. Such a policy was put into place with the Collection Development Statement for Electronic Information Resources of Penn State Harrisburg's Heindel Library.5 This type of policy can serve as a guide to support the mission of the institution and can be used to justify the selection of certain resources. Conclusion We have only scratched the surface with this study and opened doors to further investigation. The study has stimulated new ways of thinking about how collection development is done a Booth Library. From this snapshot of the full-text collection, focus was established on a few questions out of many that arise from the task of evaluating full-text databases. The title overlap aspect of this study highlighted similarities and pointed out surprising strengths. Overall, the study was very time-consuming and not practical for continuing on a regular basis. However, we hope to maintain a similar methodology to evaluate new full-text acquisitions. Generous support was provided by Circulation Services, especially Interlibrary Loan staff, to gather and organize the data for this project. It occurred to us that a major missing piece of the full-text management puzzle is lack of dialog between librarians and vendors. Without this dialog, publishers and vendors, not librarians and users, retain control over journal selection in the full-text world.6 We need accurate information about the contents of full-text products, and it is up to us to convey accurate information about the needs of our users to the vendors. 241 The study has increased our understanding of database content and subject strengths. Librarians are more confident in recommending these databases to patrons. Our discussion has shifted to include both format-based and subject-based collection management issues, and we are establishing new methodologies for future acquisitions. We have no absolute answers, but now we have some facts. Notes 1. Ruth E. Fenske and Nevin J. Mayer, "Title Coverage of Seven Indexes to Religious Periodicals," Reference and User Services Quarterly 37, no. 2 (Winter 1997): 171-189. 2. Ruth M. Orenstein, "The Fullness of Full Text," Database Searcher 5, (Sept. 1989): 21-6. 3. Joyce L. Ogbum, "The Fulltext Question: A Study in Progress" Against the Grain 9 (April 1997): 70. 4. Anna Grzeszkiewicz and A. Craig Hawbaker, "Investigating a Fulltext Journal Database: A Case of Detection;' Database 19 (Dec. 1996): 59-62. 5. Gary W. White and Gregory A. Crawford, "Developing an Electronic Information Resources Collection Development Policy," Collection Building 16, no. 2: 52-7. 6. Ogburn 70.
242
243
244
245
246
247
248 |
||||||||||||||||||
|
|
