In the stone-age of the Web, I was a supporter of the Virtual Library, or its predecessors, like the EuroGopher Subject Tree, where I incidentally became the maintainer of the History branch (see details in a German newspaper article). I helped with the general German resources catalog, and the history trees on worldwide and national level. I still have a lot of sympathy for the project, since I do not hate trees and hierarchies (as long as they are made sufferable by reasonable “see-also” cross references). Therefore, here are some thoughts how the VL might relate to RSS, Topic Maps, and folksonomies.
Here, I must confess that I am rather enthusiastic since RSS affords something that I have been waiting for since long ago.
It’s a new way of filtering links which, on one hand, has much in common with the VL idea: links filtering/recommendation is done
- in a distributed, division-of-labour manner
- on a basis of trust in the competency of the recommender.
On the other hand, however, the RSS filtering and aggregation technique offers an additional advantage: it does no longer depend on a central repository and administration for each respective subject or topic, but allows for decentralizing also alongside the information chain, i. e., intermediate relaying and brokering between supplier and consumer of information. (You might compare the former method to waiting for a small slow train that leads directly from source to destination, while the latter method is like taking the next train to the next major station, change to a fast ICE to another major relay, and finally take a local train or bus to the destination village. See also this comparison with the railroad network.)
Since the early VL times, I have always been hoping that such an organisation would emerge, since it would eventually reduce the burden of the volunteer maintainers’ efforts. That’s the reason why we put up a “German Resources” catalogue that was intended as kind-of wholesale green-grocery market where the individual specialities’ restaurants could buy their special foods (links). Except, however, for very few other countries (including Italy for a short time), this approach did not become successful.
Even worse: it was somewhat misunderstood in some German history departments because the supplier side of scholarly links (German links about history) was more popular than the consumer side (links about German history).
Now with the emerging RSS aggregation infrastructure, much of the original decentralisation and load sharing idea of the VL can be achieved and, perhaps, be improved in the long run. I am particularly happy that even smaller disciplines like this (historical ancillary sciences) start leveraging such feeds.
Superficially, this RSS technique may even appear as a threat to the VL idea, and I think it might well be that static html pages in the lower parts of the tree might gradually be replaced. However, such a mere technical shift does not affect the seasoned organisational scaffolding and the collaboration of the scholars benefiting from their mutual expertise. On the contrary: if the expected relief of the individual burdens comes true, this organisational framework might prove even more valuable.
(Other reasons for my euphorical esteem of RSS include those that also made blogs so popular: RSS is a serendipituous blend of push and pull media. And the incremental nature of the modestly sized information units, tends to mitigate the feeling of being overwhelmed by the information overload and by the task of having to create a major, completed, definite work, because one can always trust that minor errors may soon be corrected by colleagues in the social software network.)
Today’s role of VL
I think VL has never really been a competitor to search engines which have always existed (think of Veronica or Jughead). So, their rise is not necessarily a threat. Comparing to today’s other human made catalogues, however, I think there is still a competitive advantage possible in the sector of scholarly information where the others are not specialists. Even considering that “schoogle” for scholars tries to occupy this market, I suspect that there is room left for competence, since the emerging border between scholars’ and ordinary google will align more in terms of content’s source and type rather than quality, and schoogle may end up as a kind of fulltext catalog of PDFs that have to be purchased by subscribing to the same big publishers as now, while the genuine web resources (small pieces loosely joined, granularly addressible, dynamic, and current) will go a different route. Here, there are great opportunities to prove quality, not only quality in terms of scholarly excellence as schoogle, not only in terms of relevance assured by trusted recommenders and social networks (using RSS), but also quality of classification.
I think that VL’s “customers” and usage scenarios are generally not the same as those of the search engines, and I think there is no need to try to attract and serve the latter. A recently received email aptly emphasized it: the expertise conveyed through the “see-also” type hints and links, is a very distinct property. The VL’s audience does appreciate such properties that are not very popular today:
- top down browsing, gaining overview, just-in-case knowledge, learning from the see-also expertise, rather than
- relying on artificial intelligence to apply this knowledge, answering a question just-in-time, immediate problem-solving.
I believe it’s largely a matter of cognitive styles, dispositions, and preferences, whether one prefers trees or not. (Much of my weblog writing is about this fundamental difference, see for instance, Hypertext, Encyclopaedias, or Cognitive Styles, and coincidentally, the “see also” plays a central role in my thinking about this).
Topic maps are often seen from a very technical angle, with a lot of artificial intelligence flavor, to enable queries that directly answer people’s problems. I do not only think that it is much too early to expect VL maintainers to make friends with the technical rigor and schematics required by this AI approach. I also think that it is not the most important objective of the VL’s loyal customers and I doubt that new customers can be won by shifting towards this opposite side of the cognitive styles divide.
Therefore, I think the main promise of Topic Maps for the VL is not around problem-solving queries or optimized known-item searches, but rather, it will be queries that essentially instantiate sophisticated browsing, that leverage the see-also affordances. For the sake of brevity, let me try to explain what I mean in terms of Dewey rather than XTM.
For instance, legal history, in Dewey, is not positioned under 9xx.xx history, nor is it duplicated among 9xx.xx history and 34y.yy law (as the taxonomy critics would probably do, using a Unix symlink that effectively destroys the information about the expert’s decision where to primarily class it). Instead, it is classed primarily under 34y.y9xxxx, but using the xxxx digits sequence from the respective history subtree, and leveraging on the extensive “see also” information in the section headers and the instructions (“Manual”) how to primarily class blurred-border cases in doubt. And perhaps, the “Deutsches Rechtswörterbuch” project (German law etymology) is, in turn, classed under (German language) 43z.z34yy9xxxx.
Then, even if you restrict yourself on very basic Topic Map relationships like the simple traditional library classification tree relationships, you can leverage Topic Map technology to provide a very sophisticated browsing experience that, for instance, displays all parts of the tree that contain the (equivalent of) the Dewey digit sequence …yy9xxxx… in their entry, and apply this database view to a fulltext search.
I hope this rough example made clear what could be afforded for every low-level classification decision that may be much more controversely seen than the simple convention how to class legal history. And it’s often these seemingly arbitrary or willful classification decisions that, to the critics, make trees unusable, while to the fans, they convey experts’ experience.
In this fashion, I think, the competitive advantage of human expert input will best percolate to the human VL patrons. Ok, its not yet the full semantic web usage of XTM (robot patrons), but once these robot agents have apprenticed enough from their human archetypes, they will definitely appreciate this pattern of learning from the “see-also” expertise and consume it similarly in their problem-solving decisions.
So, even though the VL’s tree approach is currently very unpopular, it may be of much value for its customers who share a certain cognitive style, and, considering that wholistic right-brainedness is catching up, it may even face a great comeback, in case it succeeds in maintaining some currentness and staying abreast.
It is not new that some people have great problems with hierarchical categories and with “the librarians”. What is interesting is how these discussions have been culminating in the last couple of weeks. A rather comprehensive summary of the discussion is here, and my opinion here is simply: we need both approaches, see #52.