World Collections

Testers will notice our first cut of the “world collections” section on each of the the Item pages. This shows stuff from public collections that might be related to whatever it is you happen to be looking at. Now that we have some real content in the system we are able to see how effectively (or not) it is working.

The fundamental problem with trying to create a mashup with Museum collections is that there are very few that have exposed their collection database via an API or an OpenSearch feed. However quite a few have built some sort of mechanism to browse their collections online which means that we can query them indirectly using the Google API.

So, the way we’re doing it at the moment is this: Collectish takes all the tags for the item and uses them to run a series of Google searches. The searches are restricted to run within a the sites for a few public collections that we have chosen. If a search doesn’t return any results, Collectish shaves off the last tag and then runs the search again, repeating the process until it returns something.

Now granted it ain’t the most sophisticated algorithm around. Two obvious problems are that the order of tags on an item affects the results of the search and the searches don’t take semantics into account. Soon we’ll be building in many more sophisticated ways of hooking into data from public collections – however this will do for the moment. And of course we’ll be expanding the list of collections we’re searching as well.