The common wisdom among library staff is that "everything should be in the OPAC". This view does have virtues: If users know for sure that every resource provided by a library could be accessed through one single point of access, then that is indeed a good thing. However, real world is not that simple. World-wide web search engines have the last ten or twelve years become increasingly important access points for digital resources.
Users of web based resources expect to find what they need through search engines. To be accessible that way, material need a browsable user interface in order to be indexed. The browsable interface is also important for the users; hypertext links to other parts of the information providers web site is essential for the users ability to assess the quality of the retrieved information object.
A significant portion of our digitization effort is put into production of digital images. They have usually been stored on the server img.kb.dk. Some of these images are bundled into documents through different means, e.g., by producing multi-page PDFs. Some of these documents are made accessible by cataloging them in our OPAC, rex.kb.dk, and some through databases on base.kb.dk or resource pages on our web site, www.kb.dk. We may see this as prototypes for two strategies of publishing. The former (cataloging in the OPAC) is characterized by that it is a searchable only resource the latter is a navigable one, since most services on the our web (base.kb.dk and www.kb.dk) does have a browsable in addition to a search box. To formulate it in another way: The OPAC belongs to what one a few years ago labeled as the hidden web.
In this note I have made an attempt to compare the efficiency of the two strategies by analysis of web access log statistics, to determine which of the strategies provides the most effective access to these resources.
Most modern browsers returns information on what is called referring page. This is the URL of the page on which the browser found the retrieved link. Referring page data makes it possible to analyse linkage patterns and the ways users find their way to digital resources. For digital images, the referring page is the page in which the image is embedded, for things like HTML pages or PDFs it is the page which links to the object.
My analysis was done as follows: I gathered all access log info for the months January to October 2007 on the host img.kb.dk. I removed all accesses that could be attributed to harvesting or archiving robots by creating a list of accessing hosts that had retrieved the robots.txt file. Accesses from these hosts were then discarded. Then I created a table with two columns, the file names of the retrieved resources and the domain of server that hosted the page that linked to or embedded the objects. The number of hits could then easily be counted with using different selection or filtering criteria.
The data set contains close to 3 million hits (2843368) from users that found their way to our digital objects through linking or embedding from 1426 unique hosts world-wide. Not surprisingly a majority, about two thirds, of the hits came from links on our own sites (1889247 hits).
For answering the question I posed above, I chose to analyse digitized sheet music. A search on WWW=http* (i.e., searching for URLs with URL scheme http) in the Digital Music database within the KB catalogue http://rex.kb.dk/ shows that catalogue provides access to 3249 digital objects. I have been unable to produce a corresponding figure for what is made available through the web sites base.kb.dk and www.kb.dk.
The total number of PDF files in the music department's img.kb.dk web area is 4530. This figure is a grand total including the same objects in different resolutions, and there might also be objects yet to be published.
Of these files, only 2898 were requested at all with referring pages from the kb.dk domain. 2226 files were requested from the web sites only, 950 files from the catalogue only (out of 3249 possible). Why this is the case, I do not yet know for sure. However, on the page Danmarks Digitale Nodebibliotek there are links to "Gieddes Samling af fløjtemusik" and "Rischel & Birket-Smiths Samling af guitarmusik" in a prominent position. Most of the hits from the OPAC are for objects in these two collections. The explanatory links into the OPAC are indeed helpful; users can easier search (successfully formulate a query) if they know what to expect.
There is small overlap, 278 files are referred to from both the web and the catalogue. I will return to them below.
Assuming that all the 2226+278=2504 objects that were retrieved through links from base.kb.dk and www.kb.dk is the total number of files published via the web we may estimate that about 40% of the sheet music published by the Royal Library through the navigable and about 60% through the search only. Our digital Sheet music is thus very suitable for testing the efficiency of the search only publishing strategy.
To obtain the data I looked on objects carrying sheet music based on directory name, limiting the analysis to the mime type PDF. I only take into account referring pages within the kb.dk domain. The data is summarized in Table 1.
| Host of referring page | Number of hits | Proportion |
|---|---|---|
| base.kb.dk | 36197 | 57.6% |
| www.kb.dk | 15068 | 24.0% |
| rex.kb.dk | 11603 | 18.5% |
| total | 62868 | 100% |
What we can see is that links from base.kb.dk and www.kb.dk together attracted 51265 users and the links on rex.kb.dk only 11603, which amounts to 81% of the hits emanating from the web sites.
To be even more specific, we can extract the individual files that have been requested from both the OPAC and the web. We have found 278 objects on img.kb.dk that are linked to from both the OPAC (rex.kb.dk) and the web (base.kb.dk or www.kb.dk). If we recalculate the table above for this subset only we get:
| Host of referring page | Number of hits | Proportion |
|---|---|---|
| rex.kb.dk | 3531 | 33.1% |
| www.kb.dk | 1830 | 17.1% |
| base.kb.dk | 5301 | 49.7% |
| total | 10662 | 100% |
Within this subset of our data, the OPAC has more successfully directed users to the resources than in the overall data set. Still, within this subset with, so to speak, equal opportunity, the web sites disseminates the resources much more effectively: 66.9% of the users came that way.
We can, obviously, not tell from a web access log to what extent sheet music is printed and put on a note stand and actually performed. One usage of a URL which we can measure is to what extent other link to our resources. We got 23831 hits on this material from links on resources on other domains (see Table 3 for a summary of the domains of most important referring pages). Some of these services are harvesters (yahoo.com and google.com surely are). Among these external links, the hits coming from referring pages situated on specialized music portals of various kinds the most important source of traffic on our site.
| yahoo.com | 8870 |
| digitalguitararchive.com | 3748 |
| ami.dk | 3747 |
| google.com | 2244 |
| icking-music-archive.org | 1803 |
| helloguitar.com | 1613 |
| robertolobo.com | 450 |
| davidelsmore.com | 244 |
| delcamp.net | 141 |
| federmandolino.it | 139 |
| hellomusiczone.com | 80 |
| dcguitar.net | 74 |
| pianophilia.com | 52 |
An interesting a question is how the remote services found the resources. Obviously we will never be able to answer that question. We can, however, count how many of the count on how many of 23831 belong to the resources that are disseminated via the web services or via the OPAC or via both services. This comparison can be found in Table 4.
| URL disseminated via | Number of hits to these URLs from remote resources (referring page not in .kb.dk domain) |
|---|---|
| www.kb.dk & base.kb.dk | 17501 |
| rex.kb.dk | 8471 |
| all three | 2802 |
Finally, we know that bibliographic records are moving targets in many ways. One the ways is that records are exported from the OPAC to the national union catalogue bibliotek.dk. This has been raised as an important argument for cataloging digitized cultural heritage material in the OPAC, since (i) the union catalogue is much used and because (ii) it is collaborating with google.com.
We can at least partly refute this argument for the following reasons: The search engines are minor actors in this area (cf., Table 3). They are more important as disseminators of those of our web pages that link to the sheet music than disseminators of the digitized objects themselves. This is most likely due to the fact that these PDFs contain very little text to index. That is, the hits we get this way is more due to less than excellent preparation of our resources for harvesting, than successful export of data to google from bibliotek.dk.
As regards the importance of bibliotek.dk as disseminator of our sheet music, we have have to draw a very dissapointing conclusion: We have recorded 9 (nine) hits on sheet music from the bibliotek.dk service within this dataset.
To summarize all these data we can list the conclusions as follows. For users that came to our sheet music via servers in the kb.dk domain we draw the following conclusions:
As regards links from external resources, we can conclude that they come from all over the place, and in particular various specialized music portals. As expected, a resource disseminated via the web is more likely to attract external links than one promoted via the OPAC.
Finally, there are extremely few hits from other OPACs. The only found this year are 9 (nine) hits from bibliotek.dk.
I wouldn't claim that it is a disadvantage for libraries to use searchable databases for promoting its resources, but I have to conclude that it is close to suicidal to use them as the only access method to digital resources.