Zooming

I'm trying to decide which direction to go in developing some prototypes for textual (rather than image) ZUIs. This has two dimensions: * textual hierarchy: e.g. the Protestant Bible has 66 books grouped into clusters, like the Hebrew Bible (Old Testament): the Penteteuch, Historical Books, Wisdom Books, Major and Minor Prophets, and the New Testament (Gospels, Paul's Letters, etc.). Each book has chapters, smaller thematic units with titles ("pericopes"), paragraphs, and verses. Each of these units has a traditional reference scheme (e.g. "Psalm 23, verse 4"), and typing this into a search box is the traditional means of navigation. I'd like to instead use a zooming interface for navigating this space, by showing the relevant textual details at each level (book names when zoomed out, chapter or pericope names when zoomed in some, full text when fully zoomed in). * I'd like to integrate this with a faceted search facility based on rich metadata, allowing the user to e.g. select only pericopes where David is speaking and the word "justice" is found. I've looked a little at Silverlight/DeepZoom, but it appears to be oriented toward images. I'm looking for the capability to resize a text box based on its contents, compose them together into a larger unit (think treemaps), highlight text hits to show you where to zoom in (even if the text isn't yet readable at this zoom level), etc. Is Silverlight a good technology for this kind of interface? Would Flex be a better choice? Or are there other approaches that make more sense? I've got the text and the metadata, so i'm really looking for more general tools for building these kinds of interfaces.

Sean, Personally, I don't have much experience with Silverlight therefore I can't make a qualified statement for it. However, I know that Flash/Flex can handle reasonable amounts of text, reasonably well in a ZUI. Especially with the new Text Engine in Flash Player 10, you can do some beautiful stuff. Shameless plug: Check out the OpenZoom SDK and the sample for text rendering I've created with it. In case you need better performance, the Piccolo2D framework for Java, albeit more low-level, is probably faster than either Flash or Silverlight. In the TED Seadragon/Photosynth presentation by Blaise, I've witnessed the pretty nifty text rendering capabilities of Seadragon. However, I suppose Ian could perhaps give us a more current status on those.

Thanks for your comments Daniel. I'll take a closer look at OpenZoom. My previous prototype used Piccolo2D, but the documentation wasn't as clear (or maybe i wasn't as smart!) as i would have liked, and i'm thinking something with a little broader adoption (e.g. Flex or Silverlight) might be a more strategic choice.

The Seadragon text technology that Blaise demoed has not yet made its way into Silverlight. Ultimately the big challenge is organizing the data in such a way that you can progressively load it based on the area you are zoomed into. I don't believe Flash or Silverlight has a built-in solution for that, at least not on the scale you're talking about, so for the time being, you'll have to roll your own. Once you've got that sorted out, I'm sure either Flash or Silverlight can handle the display; for that matter, I've done some interesting experiments with zooming text in JavaScript (not published yet). A generalized tool kit like you suggest is definitely of interest to us, but we're still a ways off. In fact, this is part of why I created this group; to bring together people interested in creating such things. I'll pass this thread on to some folks who might be interested. Whichever way you end up going, I'd love to hear how it goes. Oh, on a side note, your project reminds me of this one: http://www.chrisharrison.net/projects/bibleviz/index.html

Ian: Thanks for that information: it's helpful to know what isn't available. In theory, i'm not sure progressive loading is as much of as issue for text, at least the texts i'm interested in: for example, the entire Bible is only about 600k words, so (unlike hi-res image applications) memory constraints aren't as critical. But there are other features that arise with text zooming that don't really come up with images: * i'd love to have a search interface that displayed hit highlights at zoomed-out scale but then enabled you to move in on items of interest. You can see this (without zooming) here and here. * like image thumbnails, chunks of text often have meta-data that make it reasonable to select and filter them (like the HardRock demo) * some texts have higher-level features (chapter or section headings) which are useful in their own right for navigation and drill-down, and also let you see the larger context * for larger texts, showing relative size (a la treemaps) also gives a very different perspective that's complete lost when the only view you have is close-up and linear I have seen Chris Harrison's work: it makes for interesting art, though i'm not sure it leads than much deeper understanding of the data.

Sean, We haven't come up with general purpose solutions to any of those problems, but I'd love to see what you come up with! Perhaps some of your work can be packaged for use by others in the zooming community.

Sean, We are working on Text, but Flash has some limitation which we hope ver 10 will alleviate. Maybe these examples will help you in conceptualizing your project. We put 280 HD images of art online as an exhibit. Two versions: 1) everything on one canvas, with all the text and images (approx 6000 elements in the XML), and 2) linking from one surface to another to manipulate less elements. We found a performance hit from manipulating too many elements since the XML has to be parsed every time you pan with the mouse (no pb if you only use the keyboard to navigate - I/O, tab, and arrows). We also believe there is some cursor manager pb within Flash. At least you can see Text appearing at a determined Z level. So it should answer some of your question. The fonts are loaded on demand, and all is rendered from an XML file, with CSS, and Javascript for programmability. Very close to HTML programming. 1) One canvas version: http://ak.zoomorama.com/static/onetime/artsxsw/ 2) Multiple canvas version: http://ak.zoomorama.com/static/onetime/artsxsw-split/ Does that help your thinking ?