I’m feeling guilty about how hard all the other bloggers here are working, so I thought I would try to post on this presentation on MERLOT’s new federated search technology… even though Alan is sitting next to me, probably doing the same thing. And I see D’Arcy up front, and Scott Leslie is here too. They probably even understand this stuff.
MERLOT: A focus on federated search technologies…
This approach is an attempt to address the ongoing problem of disparate collections of digital learning resources. They recognise the value to specialized collections, but it is obviously a clumsy process to be forced to search multiple databases with a single query.
Federated search approach was chosen over the “harvesting” model… They were “afraid” of harvesting. They hope that this model will preserve the unique characteristics of the smaller repositories while still offering them the opportunity to be part of the MERLOT “federation”.
Federated search client connects to partner resources: http://fedsearch.merlot.org
So far it aggregates MERLOT, EdNA Online, and SMETE.
The search results are returned along the lines of a typical MERLOT record: they contain links to peer reviews, assignments, etc…
Rankings of returns are based on where the search query appears in the record: ie) if it appears in the title, it returns higher than if it is in the description, etc… An attached peer reviews and associated metadata can serve as a “tie-breaker” to raise ranking of some resources.
This is based on existing standards that are is use, including Google API and SOAP.
Alan is asking why the search query doesn’t return a specific URL (the way Google does, for example). You therefore cannot send someone or post a link to your search results, only say “go to the search engine and enter [x]”. They seem surprised by the question.
MERLOT has created 30 RSS feeds… I guess they must be based on subject headings, and cannot be customized based on specific queries. And it seems like they haven’t built up the interfaces and tutorials for showing the community how to access those feeds. “We have the feeds, but you aren’t ready to see them.”
Alan seems pissed off. I don’t blame him. I’m not clear on why other resource collections (CAREO, MLX, even our humble little homegrown repository at UBC) can have this but the mighty MERLOT cannot. When pressed by Alan, the presenter simply replies: “we don’t have that.”
The wireless connection has gone done. They don’t seem to have a backup. There’s plenty of discomfort up front. (I hadn’t realized that they didn’t have hardlines for the presenters… damn, am I glad we took lots of screenshots for our presentation tomorrow.)
The presenters have skipped over the architecture piece explaining how they built their web services architecture. Too bad, I was kind of interested in that. Maybe they could have asked us before they moved on.
Man, I could use a drink. Is it 6:30 yet?
They acknowledge that they are having enduring problems (search syntax, results requirements)… and they aren’t even going to try to solve them… it requires a solution from the larger community. Strangely, I find this open disavowal of responsibility to be reassuring and kind of cool.
The “community charter” they need to develop involves simple standards for searching multiple collections, and a federated search framework as an implementation of those standards. It does beg questions on how a resource catalogue gets to become part of this grand federation… and what rules it will need to adhere to in the way they structure their searches.
Structured metatada is not a requirement… the organizers seem to recognise that structured metadata may be valuable, but it doesn’t make a big difference in performing effective search results. This is something of a provocative notion… the formula I’ve always accepted (somewhat uncritically) is that more metadata fields filled out increases management burden, but also improves resource discovery. But I think I agree with the presenters… the metadata might be of more use to evaluation, or application, of a resource that’s been returned… but an effective text search with limited metadata can be very useful.
OK, I’m better now.
Quote on federated search: “quick and dirty, find it and go!”
Design decisions were made with preference for simplicity… they intend to grow out their web services model… Problems of growth are desirable, as addressing them are responses to real needs from the community. Again, the strategy is passive aggressive, but I kind of like it. Perhaps that says more about me than the presentation.