Author: Mullin, Casey A
Date published: December 1, 2010
Journal code: PMUN
International Music Score Library Project/Petrucci Music Library. Project Petrucci, LLC. http://imslp.org/wiki/Main_Page (Accessed May 2010). [Requires a Web browser, Adobe Reader and an Internet connection].
The International Music Score Library Project (IMSLP), alternatively branded the Petrucci Music Library since its relaunch in 2008, is a non-profit project that operates within a simple yet formidable mission, stated prominently on its home page: "to create a virtual library containing all public domain music scores, as well as scores from composers who are willing to share their music with the world without charge." In four short years, it has progressed admirably towards this goal, becoming not only one of the largest free online collections of digitized printed music, but also one of the fastest-growing, adding on average over 2,000 scores per month.
The IMSLP gained notoriety in the music community surrounding its tumultuous early history. Founder Edward W. Guo, then an undergraduate classical composition student at the New England Con - servatory of Music, launched the site on 16 February 2006. As it gained popularity, it also caught the attention of a large European publisher, several of whose scores had been mounted on the site. After receiving two cease-and-desist letters from the publisher in 2007, Guo opted to shut down the site; as he stated in an open letter to the community, "the cease and desist letter does not call for a takedown of the entire site, but . . . I very unfortunately simply do not have the energy or money necessary to implement the terms . . . in any other way." Happily, Guo was eventually able to mitigate the complications of disparate copyright terms (as explained below), and the IMSLP was re-launched on 1 July 2008, featuring a redesigned user interface powered by MediaWiki, the interface familiar to users as that originally developed for use by Wikipedia.
SCOPE OF CONTENT
Housing over 61,000 scores (downloadable as PDF files) as of May 2010, the IMSLP rivals many brick-and-mortar music libraries in coverage. To wit, this figure is displayed prominently on the home page, alongside two other constantly-increasing figures: the number of works represented on the site (currently approaching 25,000) and the number of composers whose works are represented (nearly 3,300). The scope is broad, encompassing Western art music from all periods and in all genres. Under - standably, as the bulk of the collection has originated from users' personal collections, the nineteenth and early twentieth centuries are the best represented. However, a large contingent of living composers has begun to use IMSLP as a forum for disseminating their works, employing Creative Commons licenses. Such a forum effectively bypasses the commercial publishing apparatus, and uncovers a treasure trove of new music never before gathered in one virtual space. One young composer in particular, Eric Quezada (b. 1995), is surprisingly prolific, having uploaded over 200 works. To be sure, the editorial and vetting mechanisms of traditional publishing are also bypassed in this way; though submissions are monitored closely for adherence to copyright and/or licensing requirements, no endorsement of musical quality of any particular work is put forth by the hosts of the site. Accordingly, a "discussion" tab on each work page allows members of the site to contribute commentary and analyses of specific works. Unfortunately, I have observed that this tab is being used more for discussions of scan quality and the like. Still, the function is there for those who might wish to share their particular ideas on Beethoven's Ninth, or to opine on a composer's latest creation.
The aforementioned copyright disparities create a potentially misleading picture of score availability, as not all works are in the public domain in all geographic areas. Since IMSLP's servers are located in Canada, the baseline requirement for score submission is that the work in question be in the public domain in Canada, which observes a copyright term of the life of the author plus 50 years. Since this term is more permissive than that of the United States (as long as the life of the author plus 95 years for works published in or after 1923) and of the European Union (life plus 70 years), many scores have been uploaded to IMSLP that are not available for download either in the USA or the EU. Many such scores were the impetus for the cease-and-desist letter that led to the site's shut down in 2007 (the list cited by the complaining publisher is available at http://tinyurl.com/28d9kco). To illustrate an example, in the case of Serge Rach - maninoff, all works are freely available for users in Canada, early works (pre-1923) are available in the USA, and none are accessible in European Union countries, where his works will not enter the public domain until 2014. Conversely, works of several prominent twentieth-century composers will be rolling into the public domain in Canada or the EU in the coming years, specifically Erno? Dohnányi, Percy Grainger and Fritz Kreisler (in Canada), and Frank Bridge and Alexander von Zemlinsky (in the EU), whereas post-1923 works by these composers will be under copyright protection in the USA for many years to come. In fact, users have uploaded some scores in advance of their availability dates, so that when the works enter the public domain, said scores could be "switched on" immediately (this practice was observed recently in the case of Bohuslav Martinu° and Heitor Villa-Lobos, whose works entered the public domain in Canada on 1 January 2010). Fortunately, the Web site is IP-address aware, and each score is coded with the appropriate copyright restrictions in each of the three legal systems. The site also employs a rigorous copyright review workflow, wherein a cadre of users both well-versed in copyright laws and experts in the site's operations ensures the proper copyright coding of each score before it is made available for download. Thus, access is restricted in such a way as to preclude undue copyright infringement. Specifically, when a user clicks through to a score that is not tagged as public domain in that particular geographic area, he or she is presented with an appropriate error message (see http://tinyurl.com/25o4jr3).
What makes IMSLP most impressive, both in terms of its size and the robustness of its retrieval environment (which will be elaborated on below), is the fact that it has been built and maintained entirely by its community of users, who are not remunerated for their efforts. The site stresses the importance of user contributions on its home page, and elaborates on them on its "Contributor Portal" page. In addition to scanning and uploading the scores themselves, users are responsible for populating the work pages with historical and bibliographic metadata, verifying copyright restrictions for each score, tagging each page with appropriate categories, editing metadata to bring it into conformance with the site's style conventions, and even policing the activities of other users who violate the site's guidelines. The quality of the digital scans varies considerably, and each score can be rated for the quality of its scan. Often, low-quality scans of scores have been supplemented with higher-quality counterparts later, as users have made them available. Furthermore, users can create profile pages, where they can introduce themselves, chronicle their previous contributions to the site, and describe their future intentions. Wish lists of desired scores are also maintained, and each user can "watch" an unlimited number of pages, in order to be alerted when content changes. Finally, like Wikipedia, many pages have been translated into numerous languages. In short, this is Web-2.0-style collaboration writ large.
The score uploads originate either at the initiative of individual users (say, from their private libraries), or as part of one of the many "community projects" initiated by the site's administrators. Recent large-scale proj ects include the Orchestra Parts Project, which has made available complete parts for over 500 orchestral works, thus lending serious competition to the popular Orchestra Musician's CD-ROM Library (http://www.orchmusiclibrary.com). In addition to scores-and a considerable number of books about music-scanned and uploaded by users directly, the IMSLP contains a large number of scores "mirrored" from other online archives, for example those in the Variations Music Library at Indiana University and in the Sibley Music Library at the Eastman School of Music. Thus, the site functions as a portal to digital collections worldwide, with the potential of eventually serving as a single access point for all digitized public domain scores available on the Web.
The primary content of IMSLP is supplemented by links to commercial recordings sold by Amazon.com and to hard-copy scores sold by Sheet Music Plus and Amazon. IMSLP receives a commission on each item purchased as a result of a clickthrough, though the site disclaims any further relationship to or endorsement of the vendors (see http://imslp.org/wiki/ IMSLP:Advertising). The score links are manually added to each particular score, but the links to recordings available through Amazon seem to be automatically generated from keywords on the work page; as a result, many of these recordings do not actually contain the work in question. Despite this deficiency, the added value of recordings being made accessible on the same page as a score is significant. Non-commercial recordings (i.e., freely available elsewhere on the Web), have also begun to appear on the site as links to sound files, though these must be added manually by users. Of further note is the IMSLP's recent foray into publishing. Specifically, they have teamed up with the publisher Serenissima to issue high-quality reprint editions of certain scores available digitally on IMSLP. These scores are listed on Amazon under the new imprint Petrucci Library Press; as of May 2010, about a dozen have been published (see http://tinyurl.com/28eb8qj).
METADATA AND RETRIEVAL
IMSLP 's retrieval interface relies heavily on browsing. The front page has links allowing users to browse by composer name, time period, composer nationality, and work genre or instrumentation. These links are also present on the sidebar of every page on the site. Browsing by composer takes the user to a landing page for that composer, which may include alternate name forms, links to external biographical and other reference pages, and usercreated lists of works, which can range in robustness from hyperlinked lists in order of opus number to sortable tables (in the case of more prolific composers). In addition to providing another mode of access to a composer's oeuvre, comprehensive work lists can serve as reference sources in their own right - to aid a user in identifying a score he or she is uploading, for example. The ability to browse by composer period and nationality is a rather innovative feature, going well beyond the functionalities of most library OPACs. It should be noted that all browse menus are generated from user-supplied category tags, which may be derived from controversial or arbitrary decisions; for example, Beethoven is tagged only as a Classical composer, and Monte - verdi only as a Baroque composer. Am - biguities of nationality are similarly exacerbated, in the case of those composers who are associated with more than one country (e.g. Chopin). However, a user endeavoring to compile a program of works by, say, Catalan composers from the nineteenth century, is well-served by these browse menus. The home page also includes an automatically-generated list of the most recent scores added to the site, as well as a pane with user-suggested "featured scores," which can include notices for anything from noteworthy first edition scores, to complete works sets contained on the site.
The most innovative browse feature is the recently-developed genre and instrumentation categorization, which is still being completed (according to the site, as of 20 March 2010, just over 50 percent of work pages have been tagged according to the new system). Prior to this feature's implementation, though precise instrumentation and genre was entered in the General Information section of each work page, only a small number of high-level categories were applied as tags and thus available for browsing as genres. The newlyemployed system includes a menu of highly specific "work types" (i.e. genres), instrumentations, featured instruments, and languages. Starting from this menu, the user can build a "stack" of limiting attributes, jumping at any point to the list of work page results. The interface is not entirely intuitive, as it is not always clear which link should be clicked to achieve a certain result. To illustrate, clicking on "Rondos" results in a list of work pages. However, if "browse"-which appears in square brackets next to "Rondos"-is clicked, the result is the "Category Walker" page, which breaks down the result set by other work types, instruments, composers, etc. At this point, the "list" link next to each entry leads to the filtered result set, while the "add" and "rmv" ("remove") options filter the results without displaying the work pages (thus allowing the user to build a stack of limiting attributes before seeing the results). Oddly, clicking the limiting term itself (e.g. "For orchestra"), originates a new search on that category alone. The instructions on the "Category Walker" page are detailed and helpful, though they would be better-placed on the previous toplevel page. Furthermore, it is not clear what is meant by "featured" in the "featured instrument" categories; this could connote an instrument/voice occurring in a solo context, or merely an instrument that is part of the overall scoring of a work. A sampling of results obtained by clicking these categories seems to suggest the latter, but this is not explicit. Despite these shortcomings, the potential for very precise retrieval allowed by this granular categorization of works is promising. It should also be stressed that this categorization system is still identified as a "project" and will most likely improve over time.
In contrast to the rich browsing options available, the searching capabilities on IMSLP leave something to be desired. As in Wikipedia's interface, there is but a single search box, and no advanced or directed search options. Rather, there are four search indexes from which the user may choose: a "general" search, a "composer name" search, and two accession number searches (though the latter two seem useful only to those performing administrative tasks). Each results set is sorted into two sections: pages whose titles include a matched term, and those where text anywhere on the page matches a term. Beyond this high-level relevancy ranking (presuming a page containing a matched term in the title is more relevant than one in which a term appears just anywhere), the results seem to be sorted only in reverse chronological order, based on when the page was created. Such ordering may be efficacious in a journal article database where articles are added as they are published, but it is of questionable utility in this environment, where the date added bears little correlation with when the work was written, when the score was first published, or any other metadata meaningful in this context. Composer name searches are the most useful, as each composer's landing page has as its title the composer's name; thus, the desired page will appear at the top of the results set. Searches using the "General search" button, on the other hand, proved somewhat more problematic. A search on generic words, such as "viola sonata," retrieves a large set of relevant results, but again with no meaningful ordering. Known-item searches generally have better results. For example, "Debussy Mer" retrieves the correct work page as the first result, since the page title includes the composer's last name and the preferred title (in this case, in the original French). However, searches using popular names were of mixed success. "Beethoven moonlight" retrieves the desired work page ("Piano Sonata no.14, op. 27, no.2") as the first result, whereas "Schubert unfinished symphony" retrieves the desired work page ("Symphony no.8, D.759"), but as a "page text" match rather than a "page title" match; thus, it appears not in relevancy order, but in reverse chronological order as noted above (fortunately, in this case, it was only the second result). And more complex searches, e.g. "Beethoven sonata violin piano," fail altogether. In sum, though many searches will lead the user to the desired score, browsing is clearly the more reliable method of access on this site.
As with any retrieval environment, success or failure of searches depends most heavily on the richness and quality of the data itself. On this count, IMSLP exceeds expectations. Compared to many full-text article databases which are fraught with author name duplicates, this site circumvents this problem by having an established category page for each composer and requiring new scores to be assigned to a composer category if that composer already exists on the site. A detailed "score contribution guide" helps to minimize errors and duplication. Moreover, most data errors in new submissions are quickly fixed by other users as they appear on the site. Within each composer category, scores are organized around the concept of the musical work, much in the same way library catalogers use uniform titles to exercise authority control. Though data on IMSLP is not as tightly controlled as that which appears in a library's OPAC, the collective wisdom of its users provides a kind of "social authority control," which is quite effective in most cases; a similar phenomenon can be observed on the popular online book community LibraryThing (http://www.librarything .com/). Understandably, accurate control is more common for composers whose works bear opus or thematic index numbers, and more problematic for earlier composers, where the extent of a "work" is debatable - e.g. should a work page represent an individual motet in a collection, or the collection in its entirety? Some composer category pages show this inconsistency. Jean-Marie Leclair is one interesting example, where there are work pages both for the entire set of opus 9 violin sonatas and the famous Sarabande and Tambourin from opus 9, no. 3; presumably, the creator of the latter work page did not take the time to properly identify the origin of the excerpt, nor do the scores mounted to this page provide this information. This "social cataloging" model employed to organize the ever-growing collection of scores on IMSLP is less than perfect, to be sure, but any momentary shortcomings are mitigated by the fact that any user with expert knowledge of a certain composer or work can create an account and, once well-versed in the site's operations, perform the appropriate edits, up to and including reorganizing work pages altogether. Such is the beauty of the wiki model. And unlike the notoriously abusive and controversial practices that have occurred on Wikipedia, IMSLP has thus far not succumbed to such problems, which are arguably inherent to this type of social, user-driven environment.
The International Music Score Library Project's formidable size, high-quality metadata, innovative retrieval environment, and dynamic growth have rendered it a resource not to be overlooked. Admittedly, its existence cannot supplant the need for well-curated physical music collections, nor is its scope comprehensive enough to include recent compositions by major composers or current scholarly editions, both of which will be under copyright protection for the foreseeable future. But with its abundance of rare, out of print, and esoteric scores, it fills a sizable gap in the coverage of most academic music libraries. Furthermore, the implications for the future of music bibliography as a discipline- in the presence of such a quickly-growing and easily-accessible resource-are compelling. Most importantly, as this site is only as strong as the collective contributions of its users, music librarians might play a key role in the continued improvement of IMSLP by making their own contributions. This librarian has certainly endeavored to do so.
CASEY A. MULLIN