Document View

An argument for eliminating the DOE PubSCIENCE database is presented. There are seven reasons to eliminate PubSCIENCE. Some of them include: it has only half the claimed unique records. PubSCIENCE is built on a 960,000-record subset of the Energy Science and Technology database, which it refers to as the DOE Database. Second, there are only 34 contributing publishers. The two dozen speeches delivered since the launch of the database mention differing numbers of contributing publishers. Third, there are fewer than 1,000 journals. The PR materials refer to 1,200-1,400 journals for which the publishers supply bibliographic records.

Enlarge 200% Enlarge 400%

Enlarge 200% Enlarge 400%

Friends, information professionals, fellow taxpayers! Lend me your ears. I come to bury PubSCIENCE, not to praise it. And I'm not using that trick of Antony's, who praised Caesar in his eulogy. I do think that PubSCIENCE should be killed. As Brutus said, "Hear me for my cause" to find out why.

The Global View

Clearly, I'm in the 25-percent minority who-according to a recent Information Today, Inc. Snap Poll (http://www.kmworld .com/resources/itipolls/index.cfm?action= viewpoll&Poll_ID=16)-agrees with the idea. Library associations issued alerts that described PubSCIENCE as an "extremely important" database and distributed boilerplate letters to senators that fiercely opposed its discontinuation. It is for sure an uphill battle, but it may be worth the fight to channel the readers' energy (and the congressional appropriations) to support other, far more worthy government databases. Even if PubSCIENCE gets appropriations-or more likely, gets discontinued but not removed-it wouldn't be prudent to support it, let alone use it in lieu of the much better alternatives.

I'm not going to elaborate on what a staunch supporter I have been of the free (or "prepaid," to borrow Marylaine Block's far better term) government online information services. What I am against is earmarking appropriations for government databases that in the past have basked in both the glory and money, but that did very little to enhance their content and services.

Seven Reasons Not to Weep for PubSCIENCE

1. It has only half the claimed unique records. PubSCIENCE is built on a 960,000-- record subset of the Energy Science and Technology (ES&T) database, which it refers to as the DOE (Department of Energy) Database. The other subset of about 368,750 bibliographic records is submitted by publishers who cooperate with the DOE's Office of Scientific and Technical Information (OSTI). This would be more than 1.3 million records (as opposed to the "over 2 million" claimed in many documents). The number of unique records in PubSCIENCE, however, may hardly exceed the 1 million mark. There's an excessive proportion of duplicate and triplicate records as illustrated in Figure 1.

There are two other separate databases offered by the DOE, which include about 2.1 million and 3.7 million records from the ES&T database, respectively, that have far more comprehensive indexing and abstracting and far fewer duplicates. More about them later.

2. There are only 34 contributing publishers. The two dozen speeches delivered since the launch of the database mention differing numbers of contributing publishers. An article by OSTI's executive director in the Winter 2001 edition of Issues in Science and Technology Librarianship (http:// www.library.ucsb.edu/istl/01-winter/article2 .html) read, "More than 40 publisher agreements provide PubSCIENCE patrons the capability to search and access almost two million records in more than 1,300 journal titles of peer-reviewed scientific and technical information." The official pamphlet [DOE/OSTI-CO79 (1/01)] lists 42 contributors. There are no publisher-submitted records from Oxford University Press, the American Society of Mechanical Engineers, or the American Society of Civil Engineers, or from a few other partners that PubSCIENCE has touted for 3 years. Marcel Dekker, Inc.'s grand total contribution of three records is not exactly a sign of enthusiasm for this partnership.

3. There are fewer than 1,000 journals. The PR materials refer to 1,200-1,400 journals for which the publishers supply bibliographic records. However, there are records for less than that. The PubSCIENCE Web site (http://pubsci.osti.gov/publishers journaltitles.html) identifies 928 titles, but there are publisher-submitted records for only some of them, and for only some of the time. The American Meteorological Society simply can't submit records for the 27 journals that OSTI credits to it (as well as to the genuine contributor, thus double counting) because it publishes only 10 journals, not counting its abstracting-and-indexing publication-which it will certainly not offer for free to PubSCIENCE. There are records from only two Ziff-Davis publications while the above journal list identifies 30 titles. By the middle of fiscal year 2002, OSTI entered records for only 460 journals. If you look at Table 1, you see a significant decline for the number of publisher-contributed records, even though the fourth quarter just started.

4. It has many irrelevant journals and items. Even with stretching claims of its interdisciplinary nature, there are several journals in PubSCIENCE that have hardly anything to do with the energy research that the DOE is involved with. It seems excessive to have a dozen journals with a urology focus. Clearly, there may be some articles dealing with radiation therapy and kidney stones, but there are tens of thousands of medical articles that have nothing to do with physical science or energy, such as the 25 items about circumcision. The cooperating publishers' medical and biomedical journals represent a significant proportion of their contributions to PubSCIENCE. These are very well-covered and supremely presented by PubMed, which provides free access to MEDLINE, the National Library of Medicine's (NLM) database of more than 11 million bibliographic citations and abstracts in the life sciences. It would be much better to let the user click on a button within PubSCIENCE to also send the query to PubMed, as is practiced by the best multidisciplinary sites.

5. It contains many skeletal publisher records. Having the publishers supply records certainly cuts back on OSTI's costs of creating abstracting-and-indexing records. But as you can see in Table 1, publishers provide subject descriptors as the exception, not the rule. (Only 0.6 percent of the 368,750 records contributed by publishers have descriptors.) Even when they are provided they're not descriptors from the energy thesaurus that DOE has been using.

Abstracts appear much less often in the publisher-contributed subset than in the DOE-created subset. Yes, there are links to the free abstracts at the publisher site, but the abstracts are obviously not searched except when they are physically included in the PubSCIENCE records. OSTI does nothing with records received from publishers; it just dumps them into the database on an "as is" basis and creates an automatic index. In contrast, the information specialists at NLM enhance such records with value-added information, except for those that turn out to be of marginal interest for the scope of PubMed.

6. It contains disappointing links. While it's true that all of the publisher-supplied records have a link, the implementation of the links varies widely. For some publishers, like Institute of Physics Publishing, the links are perfect. Although its records in PubSCIENCE have abstracts, it's better to link to the record at the publisher's site for the many extra features available there. The links to Blackwell are sometimes problematic, as the PubSCIENCE records often explicitly promise abstracts at the publisher site. However, when you get there you realize that there is no abstract. Often, the link is to the publisher's home page, not to the article. (This is true even when the unambiguous Digital Object Identifier is there in the PubSCIENCE record, but OSTI cannot or does not want to make a link of it.) You're on your own if you want to find the journal and the article, thus making you feel like the stork at a dinner hosted by the fox.

The real problem with PubSCIENCE is that the tales about its virtues were told by so many (including us, the reviewers) that they were assumed to be true. One really couldn't hope for improvement after hearing an OSTI executive say the following at a conference in 2000: "PubSCIENCE now covers 1,032 journals with 26 participating publishers, as well as 1.7 million journal citations. Now compare that to PubMed. PubMed only has about 600 journal titles." Even if he meant 600 titles with linkouts, it was like Dan Quayle comparing himself to President Kennedy. If you look at the same record in PubSCIENCE (Figure 2) and PubMed (Figure 3), you get the equivalent of Senator Bentsen's reply to that delusional comparison.

7. There is greener grass elsewhere. The irony of the PubSCIENCE case is that the much better free subsets of the ES&T database were practically shunned. If you never heard about them, you're not alone. While the http://pubsci.osti.gov site has nearly 17,700 links, the http://www.osti.gov/energycitations site (which has the 2-million record version of ES&T) has merely 586 links. There is an even lower number of links to the largest version at https://www.osti.gov/doeecd. This is understandable as the site is meant for DOE staff and contractors, although in reality this does not seem to be a prerequisite for accessing it. There was no warning, let alone authentication, when I got there.

Conducting a search on either site as well as on Scirus-which has finally set up an outstanding and free abstracting-and-indexing database of a couple million scholarly articles with nifty software-will satisfy most of your energy-related scholarly information needs.

The $500,000 Question

Many articles about the congressional appropriations debate last year and this year (such as the one in the Chronicle of Higher Education at http://chronicle.com/fre2001/ 08/2001080901t.htm) mention that the annual appropriation for PubSCIENCE alone was $500,000. Even a tiny fraction of that appropriation would be excessive considering the work it required to incorporate the records provided by the publishers into the subset of the ES&T database, which has been built quite competently from other appropriations.

It would have been (and still could be) useful to enhance with article-specific links those ES&T records that were created by OSTI information professionals who assigned descriptors, subject codes, and other data elements to them. It's still not too late to do so, salvaging what's useful in PubSCIENCE before abandoning it, then focusing the energy, money, and publicity on the Energy Citations Database. Check out ECD before you follow the prominent links to Scirus and Infotrieve, which seem to me like reverse lobbying. Perhaps it's a government employee's individual bid for a golden parachute, rather than endorsed DOE policy.

Enlarge 200% Enlarge 400%

Enlarge 200% Enlarge 400%

I liked PubSCIENCE at first sight in 1999, when there were no alternatives for free energy science literature search. Why did I turn against it? Because as time went by and I saw its idle claims, I grew increasingly unhappy with it. I can't even respond as Brutus did when he was asked why he rose against Caesar. It was not that he "loved Caesar less, but that [he] loved Rome more." In light of my words above, it may not mean much. But I love my profession much more than I ever liked PubSCIENCE. I urge you to fight for more appropriations for high-quality free information services, but don't put your support behind PubSCIENCE.