Harvest

This week we are exploring harvesting metadata across collections with the goal of forming useful strong federated searches. I won't go into the difficulties of my own data harvesting experiences except to say; thank goodness it wasn't the corn crop or we'd all starve. That isn't said to diminish the usefulness of harvesting metadata and the concept of federated searches but only to state that I had trouble with the PKP harvester software itself. A lot of trouble.

Thank goodness for service providers, sites harvesting and storing metadata from various collections. I looked at four of these sites.

1. Sheet Music Consortium – The consortium is a partnership between John Hopkins University, UCLA, University of Indiana and Duke University Library.

2. Perseus- http://www.perseus.tufts.edu/ A digital library of resources for the study of the humanities

3. Geo-Leo http://www.geo-leo.de/geoleo/www-docs/?language=en collects records in the subject areas Geology, Mineralogy, Petrology, Soil Science, Mining, Geophysics, Geography and Thematic Maps.

4. Citebase http://www.citebase.org/search "Citebase contains articles from physics, math, information science, and (published only) biomedical papers>"

A useful federated collection is based on records where there exists a commonality in the metadata and that enhances your research in some way. Each of these services fulfilled that need for the most part. Perseus returned over 1,000 hits on the search "railroad stations United States." This can be attributed to the Library of Congress being one of the sites searched. Also, Perseus returned irrelevant records including a sheep shearing station in Australia. The other sites returned data as I expected. Oh yes, one other reason why harvesting is good, there is the fact that you are only doing one search instead of searching individual sites.

The University of Michigan's mega repository Oaister http://www.oaister.org/ indexes about 15M records from nearly 1000 contributors. This is one of those cases where size matters and helps the cause. The advantage is you have the potential of more sources returning more useful records making for a stronger federated search Should this be your only source for information? Probably not, your research plan should include a look in more then one repository.

The Subjects

A question peered over the top of the laptop as I was entering my photo collection into Eprints. Which came first, the subject taxonomy or the collection? Do you fit the collection to the taxonomy or taxonomy to the collection? It depends on the type of collection, I think. In the case of my collection of photos from a recent trip to Alaska. I shaped the taxonomy to fit the collection by developing a series of parent child relationships. For example:

A: National Parks
a1: Glacier Bay National Park
B: States
b1: Alaska
b2Washington

This hierarchy proved useful in cataloguing this collection, developing relationships between the photos and increases the searchability of the collection.

The challenges were where to draw the line between parent and child and develop terms that might be consistent with similar collections. Juneau is both a child and a parent.

C: Cities
c1: Sitka
c2: Juneau
c2a:Mt. Roberts
c2b: Gastineau Channel

I wasn't sure how to classify some terms. Should Mt. Roberts be a child of a different parent such as mountains? Consistency across repositories at my level of cataloguing is an ideal that I attempt to attain by drilling down from the general to the specific. In that way queries of my collection will return a useful result. That’s what it's all about, isn't it?

Pour into Drupal

I have a friend who has an old Oldsmobile 88, from the mid 50s. He'd be out there every evening, in the garage, tuning it, polishing it, just fooling with it. It never worked totally right. Drupal is like that for me. I'm tweaking it and looking for fixes to correct those messages with the pinkish background. It works but has its problems. OK, I've only been at it for six days or so and have been able to enter content and metadata. It is searchable. It works. My friend was able to drive around in his Olds but it was soon back in the garage for more tweaking and tuning.

Drupal is open source and there are many people contributing various modules (code) to add to the functionality. There are modules a plenty, for almost anything you might want to add. As a class (see previous entries) we added a number of modules to the Drupal core) including an image module and a WYSIWYG editor, both working well and a category browser which for me never worked, evidently, there is a reported bug. With the image module, I was surprised to find it only supports jpeg, gif and png files in the "attach image module". If you want to add a tiff or a bmp then you need to do it in using the more conventional file attachment.

I added three modules on my own. The first is Lightbox2, a java script which overlay images on a page. It was a method to view photos in my collection. I still haven't gotten it to work to my satisfaction. The second module is Flash Video, a very powerful solution for playing video clips. One of its more interesting features is that it converts the original format to flv so it will be able to be played in a flash player. I've been able to upload an mpg file and have it playback but in QuickTime. Needs some tweaking! The third module, Node Map, is a tool that relates a node to a location on a map using either Yahoo or Google maps. It offers many uses to visitors to my collection, primarily attaching photo to a location. It also has its bugs. When I enter a location into the map I receive a location does not exist message.

I won't be in the garage, instead I'll be at the computer looking for solutions and tweaking.