A few of you may have noticed that Facetmap has added new functions, in fits and starts, over the years. My first draft of the previous sentence said that "Facetmap has improved," but it's hard to claim improvement when your new functions are all hidden, unexposed by the API (or tacked onto the API as a hack). So it's time for an overhaul -- a new Facetmap software architecture.

I've posted a draft of the Facetmap Java API, version 2. We removed the clutter and replaced it with new types of facets, new functionality, and support for new file types. Hopefully it better conveys the model of classifying resources in multiple facets, and browsing through selections based on those facets.

Okay, well, also, you can write and use additional types of facets, like dates and maps and text-search keywords. And, from the top of our request list, you can assign multiple headings from one facet to a resource -- Facetmap will automatically make a TagFacet for that heading and group them accordingly (thus staying within the bounds of Strict Faceted Classification).

XML-level interactions with Facetmap won't be affected; this is just a change to the Java API. It's a major rewrite of the API, though, and therefore it's not entirely compatible with version 1 of the Java API. Code written against v1 will probably not link against v2 without some modification.

But this new spec is still a draft; we want your comments on it. If you're a software developer, we want to know if you could program against it to build what you want. If you design webpages, we want to know if this API gives you the data you need. Please send me any questions/feedback at trav@facetmap.com .

Here's an overview of what we think are the significant improvements:

• Mutability removed from core API; the resulting model encourages implementors to construct new Facetmaps instead of modifying/compiling existing ones. Lower-level construction of Facetmap objects has also been vastly simplified.
• Selection handles more than one heading per facet, so that a user can specify "select resources matching Heading A1 OR Heading A2 in Facet A". This doesn't mean a classifier can assign more than one heading per facet to a resource (but that can be done via TagGroup; see below).
• Configurations and adjustments are now controlled universally by property settings (string-based key-value pairs), via Java's Properties class. Selection behavior is now controlled by properties, instead of hardcoded into methods in the spec. The SQL package's JDBC properties, database column names, etc. are managed by properties. Facetmap implementations now choose which properties they support, and can define new properties.
• TagGroup introduced to handle tagging within the strict faceted classification model. A TagGroup is an intuitive grouping of tags, where each tag can independenly be assigned to a Resource. A TagGroup is therefore not technically a Facet, but it's a useful concept in practice, so Facetmap provides special handling for it.
• Selection logic split up and helpers added; easier to implement Selection.

By far the most common question asked of Facetmap is, "Why can't I assign multiple headings, from a single facet, to one of my resources? The earliest recorded conversation I can remember is the occurrence exclusivity thread on the then-great XFML newsgroup. I've since had the discussion with countless customers and scholars, and I'm ornery enough to stick to my guns of facet theory, acknowledging and ignoring market demand to the contrary.

These days my answer to the question is that it's an attempt to differentiate the faceted classification model from the tagging model. Tags (a la del.icio.us et. al.) can be grouped, providing much the same browsing interface that Facetmap does, but without the additional guarantee that the resource's facet is definitively classified. Since you have the option of a tagging model or a faceted classification model, I don't see any reason that the two models should try to be the same.

(That said, I fully expect to build tagging support into Facetmap at some point -- right around the same time we add arbitrary keyword-searching -- but when we do that, it won't be handled in software as faceted classification, because it is different.)

Anyway, my body of thought on the matter is detailed in the paper "Strict Faceted Classification", for which I've posted a teaser in the Publications section of this site. I'll also be presenting it at this year's IA Summit. If you're coming to the summit, let me know!

FacetMap continues to bring you innovations in faceted classification. There's a certain type of facet which almost everyone would like to use, but which no one has offered -- until now. The new Spectrum facet type allows your users to navigate numerical data, by specifying their own range of numbers instead of picking from a list of arbitrarily predefined ranges. One look at our demo and you'll understand.

New Audience

Faceted classification is applied increasingly often to information systems. We invite the uninitiated to explore the capabilities of faceted classification here on the site. Learn what it is, and why information professionals around the world are starting to organize their data with facets.

New Software

FacetMap isn't just a technology demo anymore. Now you can download FacetMap software and enable faceted browsing on your own system. Design your own display pages, plug in your own pre-existing Resource objects, and choose the FacetMap engine that best suits your system setup.

It was an okay episode, but I really like the idea of all these legendary power figures coming together and focusing on YOU. This is what information browsing should be like: powerful structures that, as far as you can tell, were designed specifically for YOU.

FacetMap's new advance helps sustain the illusion. By supporting the "href" attribute (see the DTD), it lets you map your resources to facet taxonomies anywhere on the net, just as if you'd written them into your facetmap file. What's more, you can make your own improvements on the imported facet taxonomy.

Also, if you have a facetmap available online, instead of uploading it to our demo site, you can just provide us with a link, and we'll stay up to date. Enjoy!

So there's this commercial on TV. I see it whenever I get the chance to watch a football game. A guy says, "I want a used car." He's in the Matrix or something, so a bunch of cars zoom up around him. He says, in turn, "a convertible", "red", and "with 150,000 miles". At each selection, zoom! different cars show up, and he ends up with the perfect car.

Great facet work. He's specified the facets he wants, and didn't have to worry about all the facets that are, for him, irrelevant. He didn't even have to worry about the make of the car that he drives away.

The thing is, the autotrader.com site hasn't been watching their own commercials. Try to find a car there, and you can't do a damn thing until you select the make of the car you want. The guy in the ad would have to do the same search 47 times, one for each car manufacturer.

AutoTrader, or anyone else who wants to see a better solution, FacetMap is for you.

But sometimes I look toward the inevitable future task of handling controlled vocabularies -- and handling them in a sustainable way. It's no small task to write, let alone maintain, a group of facet taxonomies for an everyday subject (ask Christina and Karl).

Universal Faceted Classification Of Everything is possible -- it just can't be done by a single entity, it must be distributed across the internet. (Tim Berners-Lee, story goes, originally conceived of the Web itself as a central database of hyperlinks, till he realized that would doom his system to non-scalable insignificance.) And so:

FacetMap is taking steps in several directions at once to extend faceted classification into an extensible, reusable, global resource. One step is the ability to link your facet structures or individual facets with those of others, so you can use a prewritten taxonomy, or search for other people's resources that fit into your facets.

Another step is the recognition of different file formats. FacetMap already understands XFML files, and we'll be adapting more generic descriptor languages as time goes on.

Finally, we'll be freely distributing some of the source code, which means a solid API for developers to extend the app for their own needs -- read a new file format, alter the engine to fit their system, or write a brand new feature. User interface developers with JSP skills can benefit from this too, and write their own display screens to navigate FacetMaps (as you've probably noticed, aesthetic navigation design would be an improvement to our demos. If you've got one, we want to know!)

More as it happens.

(One consensual complaint about facets goes like this: "Users don't want to learn new complex systems." This was the axiom that inspired the original FacetMap browsing interface, which is the same old system -- to the user, that is, but not to its designer.)

But as to the myriad of articles touting facets: They may tempt you, but then they leave you hanging. The examples are good, but where do I start if I want to design a facet system of my own? Is there a common language or data model? And do you really expect my client to write this software from scratch?

And now the plug paragraph. We're here to get you started. FacetMap is the first public tool that realizes your data, that maps a bunch of resources to a bunch of facets so anyone can browse your system. So we hope you like it. You can express your facet map in plain text, or in an XML format we've designed. And soon we'll support other formats that share our purpose. Expect to see support of XFML soon, and we're working on just the right way to adapt RDF.

More as it happens.

Now there is FacetMap, a solution so simple that we can take any metadata you've got, and turn it into a browsing system right here on our servers. Every page is generated when you request it, in order to display the efficiency of our unique technology.