...all about showing you nice things from various cultural heritage organisations that are related to... well, whatever you're currently looking at. Try it out with the form on the right hand side, then if you like it try the bookmarklet and feedback.
...is that some text is analysed and this analysis is used to pull out stuff from various museums and related web resources. The result should be a nice carousel of intriguing images, all linked to further information.
One way to use it is as a bookmarklet, which means you can try it on pretty much any web page you visit. All you do is select a bit of the text and click on the bookmarklet; if all goes well, you'll get a carousel full of interesting things and perhaps a few surprises too. The bit that makes it all work can also do other things besides the bookmarklet, find out more here.
Text is sent off to a content analysis service and with any luck this will send back a bunch of words or phrases that the service thinks are especially important or somehow characterise the original text. But that's not the end. The first five of these are then sent off, one after the other, to another web service, where they're used to search for cultural treasures/detritus (depending upon your perspective). We put a selection of these (at present no more than 5 for each term) into the carousel.
There are multiple web services ("APIs") for both of these operations, and when the carousel shows you get to choose different ones if you so wish. As at 23/10/20 you can choose from the following APIs:
Content for display:
Please note that some of these are very slow from where I'm sitting, and there are problems loading some images e.g. in Europeana's carousel.
I welcome any tips for other APIs to include here. One suggestion from @portableant (Dan Pett at the Portable Antiquities Scheme) was to offer some of the services that extract geographical information. We could then search for objects using place names as a keyword or in some APIs by using coordinates to search.
I started off with a mashup to test the alpha Europeana API, using Zemanta for content analysis. The main idea was to provide content that is contextually-relevant (or sometimes not!) for the page you're looking at, and a bookmarklet is a really portable way to do this, since you can take it to pretty much whatever website you visit, so I worked on that but making it flexible enough to do other things too (see next section). I also liked the strongly serendipitous aspect, which is half the fun: don't expect it all to make sense! This is partly down to the limitations of all the services, that don't always do the most intelligent things with search terms, but it's often the "near misses" that are most interesting.