Reference Reviews

Péter's Digital Reference Shelf

February 2007

Title: Choice Online Reviews
Publisher: Association of College & Research Libraries
URL: http://www.cro2.org
Cost: from $135 to $360 see options
Tested: January 25 – February 15, 2007

The Context

The U.S. may not be the top country in terms of new books published per year, or the number of book titles published per 100,000 capita, but I bet (as there are no UNESCO or ABA statistics) that it is the leader in the number of digital and print publications dedicated to book reviews, and in total book reviews published per year. For this column alone I have reviewed some of the most important digital book review collections, such as the New York Times Book Reviews Archive, the digital version of Publishers Weekly or Booklist Online. I have not reviewed the Kirkus Reviews database, but I plan to do so. This love for books and book reviews is mirrored by the very large-scale digitization projects by Internet Archive, Project Gutenberg, MSN and Google, whose Google Book Search database has improved a lot since my review a few months ago. Amazon keeps enhancing its wonderful book store in many ways at a pace that I can hardly keep up with in this column, including the incorporation of well over a hundred thousand reviews from journals and newspapers. Thomson Gale and H.W. Wilson have been providing indexes to literally millions of book reviews for decades. It tells you something that in the new Book Review Index Plus, Gale has more than 5.5. million records about book reviews, and about 824,680 have full-text according to my current test (much more than the 634,000 mentioned in the PR material. Book Review Digest Plus of H.W. Wilson, which was my pick earlier in Online, has excerpts of more than 1.3 million reviews with 10% of them in full-text. Academic Search Premier of EBSCO has "only" 1.6 million records for book reviews, but 960,000 of them are available in full text. ProQuest's Research Libraries database has 1.7 million records for book reviews, and nearly 1.4 million of them are available in full text.

The Content

Of course, the volume of books published does not guarantee quality, and certainly has included many — if not a million — little and not so little pieces that not even the mother of the authors could love, or ones that become bestsellers for a week not that much for their literary or content value but more by virtue of a "You will llllove that book" reference by a celebrity of the moment. This is not a problem for CHOICE Online Reviews (COR) and its users as COR focuses on books for academic libraries. Many of its books reviewed are also important for large public libraries with significant non-fiction collections.

Database size and composition

The print version of CHOICE has been published for 44 years, while the online version goes back to 1988. This means that there are more than 100,000 book reviews not counting the essays and other editorial materials.

It is interesting to see that when I reviewed the online version six years ago in this column, it had only 70,000 items, so it has increased by almost 50%. This puts it into the same league, by size, as Booklist and Publishers Weekly. The big difference is that there are no reviews for fiction, and real or pseudo documentary type books in COR.

There are reviews for 6,834 books with a publication (more specifically copyright) year of 2005. Of those, the titles on social and behavioral sciences received the most reviews, almost 3,000, followed by 2,210 books in the Humanities, 1,165 in Science & Technology and 536 in Reference books.

The profile by disciplinary groups, shown in a chart that I felt compelled to create from my test results, is similar for the 2000-2005 period and for 1995 (for a little older test window). True, there was some decline in the percentage of books in the Reference and Social & Behavioral sciences categories between 1995 and 2005, while the share of books in Humanities and Science & Technology grew. I also tested the data for 2006, but it is too early to judge it as only about 50% of the books expected to be reviewed from 2006 had reviews as of mid-February, 2007. As shown in another chart, the rise of reviews for Reference publications grew noticeably in the partial 2006 subset to 9.15% from the previous year's decline to 7.3%.

Other Features

Beyond the reviews there are editorials, letters to the editor, corrections and hot topics which are very few in numbers, but are important. The list of hot topics should have the topic title right in the list, so that without clicking, the users would know that the focus of the 2007 February hot topic column is Food Safety.

There are only 13 units in the forthcoming titles section, but they are very useful to learn about upcoming publications and their prices, such as these ready reference publications, in planning your spending from the budget whose buying power constantly keeps shrinking.

On the surface, the bibliographic essays with 17 items in COR may seem to be negligible components by size, but their content is top notch. Librarians who don't suffer from attention deficit disorder and who don't judge a book by its cover and/or the often corny decorative star symbols on it will enjoy reading the essays, which exude professionalism and composition abilities. They are substantial (lengthy in the eye of undergraduate and even many graduate students).

They are enhanced by a list of references to the books mentioned in the essay and/or are related to the topic of it. For example, the 2007 February bibliographic essay is about Masculinities and Men's Studies. It is long indeed, more than 5,000 words, but it is well structured, and is enhanced by 57 book references, and one Web site reference. The items that were reviewed in CHOICE (and from 1988 in CRO) are marked with the accession number. In this case about 40% of the books, 23 items in the reference list, were reviewed.

It is interesting to think about why some of the others published after 1987 may not have been reviewed, but is more important to link the entries that were reviewed to the reviews, which is not yet possible. This is one of the deficiencies of CRO, but it affects only a few hundred items. All the book entries should have a link to Amazon (which makes linking child's play). Chances are good that the books would have substantial complementary and complimentary content, such as Table of Contents, Excerpt, list of references cited by the book, and list of books which cite the book at hand. A small test that I did before I submitted the manuscript confirmed this assumption.

See more about this important issue in the software section, where I discuss what I found for the list of the books in the cited references of the 2007 February bibliographic essay. This would have been even more obvious if I had done it for the 2006 December bibliographic essay about Muslim Extremists and Violence. The reference list of that essay is dominated by books published in the past 2-3 years, making it much more likely to have a larger percent of precious content enrichment materials in Amazon – but I had a deadline for this column, so I restricted my test to the reference list of one essay.

The Reviews

The reviews represent the meat of this database, of course. There are 9,331 reviews for digital sources (referred to as electronic resources), the majority (8,701) of them Web sites like this one about the Essential Science Indicators. The rest are DVD and other digital media, but about 90% of the reviews are about books.

The book reviews are short and to the point, mentioning adequacy for primary audience level (more about this later), and special fortes of the book, such as exhaustive bibliographies , I particularly like that many reviewers put the title being reviewed in context by referring to one or more books on the subject, pointing out the strengths and weaknesses, the contrasting viewpoint, or otherwise complementary nature of the book reviewed. The reviews are signed and even the affiliations of the reviewers are noted – a useful hint, clue, and trait in the happy world of commentaries and opinions of anonymous people forming social networks.

It is an important feature of the reviews that they define the primary target audience(s) or readership levels of the books, and the review grades. These have been part of the text, but lately they have been specifically added to each record functionally acting as a separate subfield. These can be used as filters for the search to focus on books, most appropriate, say, for graduate students or faculty, and earning an essential, or highly recommended.

Here comes the problem. Grades have been used as a distinct field only from 2003. This special marking may have started in 2002 because only 40% of the reviews of that year have grade equivalents. As you can see from this chart, about the distribution of reviews by grade, from 2003 onward all the records have grade equivalent, and before 2002 none.

The problem is that users don't have any idea about it, they just want to find some of the essential or highly recommended books in a subject field and use the grade filter. It is fine for the 2003-2006 window. The 0 results for the pre -2002 period would probably alert the users that something is wrong, but the year 2002 results and any results for a query that includes pre-2003 grade criteria the results will be distorted because 60% of the 2002 and before reviews are implicitly excluded in such a query. It aggravates the problem that the otherwise good and sincere Help file claims that every record has a recommendation level. This is just not true. The situation is somewhat different, but not identical with the target audience. See the software section for a temporary solution.

The Software

The original software has been replaced by a more pleasant and convenient one. What was OK in the mid-1990s looked clunky by the mid-2000s just as a father's wireless phone from that era looks to his teenage child today.

It offers searching by 11 criteria – practically by every data element, such author of the book and author of the review, words in the title, ISBN, publisher, etc. In addition, for filtering the search, additional data elements can be used, such as CHOICE Subject Headings, copyright years, grades, readership level, etc. Some of these can be also browsed through pull down menus, others are shown as check-boxes. The LC classification ranges are not browsable, neither are author and publishers' names.

The queries are echoed in the result page, except when the result is 0. That is when you would need to know if it is because you mistyped a word or used a wrong search filter. You can't, so you must type in the query again.

I discussed the problem with the review grades earlier. The problem with the readership level is somewhat different. Here the help file is accurate, saying that records may have been assigned readership levels. My tests kept showing weird results when I used this filter. For example, there are reviews for 6,761 books of 2003. Of these, 6,761 are for general readers, and 6,761 for graduate students, equally absurd results. Apparently the software goes berserk. The problem with the grades could be solved by offering checkboxes for 2003 and the following years only. The readership level problem must be addressed in a different way.

Coming back to the good features, it is very nice and uncommon that you can sort by 10 criteria, and at two levels, such as first by author, and within author by publication year. Intelligently enough, there are ascending and descending options – exactly for the above scenario. The last time I saw similar sorting options was in Ovid and Dialog. The result can be exported in a couple of formats, including export format for Excel, excellent for those who want to evaluate a result list by say, composition, in terms of publishers, publication year, disciplinary groups, CHOICE subject headings. The number of records (25), which can be displayed on a single page then downloaded, should be increased and made more simple in one fell swoop. You may add hits to a list beyond 25, but it is less convenient.

It is laudable and appealing that there are ISBN-links from the book reviews to OCLC's WorldCat, but there are no links to the books mentioned in the reviews, in the bibliographic essays and in the lists of Outstanding Academic Titles - even when they were reviewed in CHOICE after 1988, i.e. are present in CRO. The same ISBN-query link should be sent also to the user's preferred library to see if it has the book, and if it is available. The same ISBN link should be used for Amazon to find many worthy value added information. Actually, instead of sending the users there, the data from Amazon should be brought to the users, meshing them up with the CRO records.

These Amazon data are perfect targets for Web 2.0 applications featuring mesh-ups. Such features beg for an AJAX solution also in the list of Outstanding Academic Titles which are awarded every year, and help the acquisitions librarians, as well as the end-users in finding the best items (in the judgment of competent reviewers) for their money.

To illustrate my point I checked in Amazon the 57 books listed as references in the February, 2007 bibliographic essay, to find out what value added-information is available for them. The results confirm the utility of such a link.

Of the 57 reviewed books, all of them were found. Any edition with exact title and author match was accepted, irrrspective of the year of publication. Twenty-four books had a total of 65 reviews, mostly from Library Journal, Publishers Weekly, Kirkus Reviews and Booklist. Three of the books were reviewed by 8 sources. The Search Inside The Book feature, Table of Contents and predefined excerpts were available for 41 items. Text statistics, such as most often occurring hundred words were available for 37 books. 26 book records had list of 2,778 cited references, and 40 were cited by 1,775 books – not all unique titles of course. This was way above my expectations, and shows how useful such a mesh-up could be.

CHOICE Reviews Online differs in one important aspect from the other journals to book reviews, such as Publishers Weekly, Booklist and Kirkus Reviews. Their reviews are available in full-text format in a variety of aggregators' databases, such as Expanded Academic of Gale, Academic Search Premier of EBSCO. A large part of Publishers' Weekly is also available free of charge from the site of the publisher. Reviews from these sources are also available free of charge through Amazon, but not from CHOICE (except for very short snippets). Master File Premier does not cover it, Expanded Academic has indexing records for 280,000 CHOICE records, Book Review Digest has substantial excerpts, but only ProQuest has the full text of CHOICE Reviews for about 114,000 items.

When the glitches are fixed and the links discussed above are implemented, the software would match the excellence of content, and CRO will have a promising future. It is worth the licensing investment for optimizing the book purchase activity, especially when the suggested mesh-up with Amazon data provides immensely useful data about the clout of books reviewed in CRO.

— Péter Jacsó