jump to navigation

small is beautiful February 26, 2007

Posted by dfhuynh in research, semantic web.
1 comment so far

(The title of this post has been stolen from E.F. Schumacher’s infamous book, “small is beautiful: economics as if people mattered.”)

MIT has a huge Department of Architecture and I’ve been flirting with it, trying to do a minor in that field. I was able to take one course called 4.101 – Experience Architecture Studio in Fall 2004 and I absolutely loved it. In that course, I learned to think about designing at the human scale.

Most of adult human beings scale from, say, 4′ 6″ to 6′ 6″ and all of the artifacts we have ever made have dimensions somehow related to that scale. Our houses are not 3′ tall nor 10′ tall. Windows rarely start at 1′ above the floor or 6′ above. Hallways are not 2′ wide. Necklaces do not weigh 10 lbs. Eyeglasses frames don’t measure 1′ across. Etc. etc. Imagine living in the land of giants where whenever you want to sit down on a chair you have to climb on it. Or imagine that furniture pieces are designed primarily to fit into shipment containers or delivery trucks rather than to fit our needs.

Of course, there are ceilings way higher than 10′ above the floor, like in cathedrals. These supersize artifacts are intended to awe their viewers. Their dimensions were intentionally designed to contrast with the human scale. And in that, their dimensions are in fact related to the human scale.

It’s easy for architects to be mindful of the human scale (although not all of them are). It is harder for computer scientists to be mindful of the human scale, which is often the extent to which human beings can deal with information–perceiving it, understanding it, and making more information.

Talking about the human scale in the context of the Web and the Semantic Web is interesting to me. I find certain cornerstones of the Semantic Web incongruent with the human scale. Take URIs (uniform resource identifiers) as an example. Each URI is a string that identifies a single something: a concrete object like the book called “Beautiful Evidence” by Edward Tufte that I bought in May 2006; a less concrete object like the iced mocha that I drank this morning; an aggregation of many somethings like this newly wed couple or the army of Sparta; a concept like love or immortality; etc. etc.

URIs is a brilliant concept. And URLs, too. There is on the rise the URL-speaking society in which two people can communicate just by sending each other URLs, each of which resolves to the same web page (mostly).

But there is a big difference between URLs and URIs. Whereas URLs are mostly read-only (and copied and pasted around), URIs demand writing, at least at this stage of the Semantic Web. People are required to make up URIs to talk about things. Making up URIs is like making up file names: both are pretty cognitively demanding. Quick, make up a URI for that iced mocha I drank this morning… Hmm, maybe


Right. Yes, I surely want to do that for every single thing I want to talk about. Wrong, you say–you don’t have to do that much work–just generate a number:


OK, but is that something most people would want to do? Have you ever seen anyone naming their files using that method just to save time and efforts? Let the machine generate that random number, you say. Then, show me a Semantic Web application that can generate a URI for the iced mocha I drank this morning. Not just generate–mind you, it has to let me interact with that URI from now on, hiding that URI behind friendly UI.

This little difference between URLs and URIs–the former are read-mainly while the latter are write-mainly–can be hard to understand. URLs are bound to the physical world. Almost every URL is bound to some physical hard disk sectors and to “write” a URL, you write onto those sectors through some rather physical actions. URIs, on the other hand, are created in the abstract. I believe it is this complete detachment from the physical world that stretches URIs beyond the human scale.

What do I mean by beyond the human scale? Well, making up a URI is pretty demanding because you have to make sure that the URI is unique all over the world, and across time. You have to see globally, perhaps universally, and well into the unforeseeable future. A person who is tasked with making up a URI simply wants to say something, and putting this burden of universality on her at that moment would stop her from saying anything at all. Another outcome is that she completely ignores the universality requirement and just makes some clearly-nonunique URI. It’s a little bit telling newly wed couples about global overpopulation. Good long-term concern but wrong time to bring it up. 🙂

So yes, after all this ranting, the point is that while the vision of the Semantic Web is good and all, we need to think about it in term of the human scale–re-thinking, in fact, every cornerstone that it depends on to see how it fits into our mortal minds.

We need to think small to achieve big. Think local to achieve global.

This doesn’t mean that we should abandon URIs altogether. It just means that we ought to start working from the user’s end, not from our fantasy’s end, for any hope of our vision to become true.

Semantic Web researchers have started to “come over to the dark side” after my talk at ISWC 2005, buying into this idea of designing for users rather than for technologies’ sake. And while they have talked a lot about that, they haven’t done much. Look at most things called “Semantic Web browsers” out there–you still see URIs all over their interfaces. It’s sickening. Yes, think of furniture pieces designed primarily to fit into shipping containers and delivery trucks rather than to fit our physical bodies, because those interfaces are primarily designed for machines, not for humans.


Calling all EXHIBITionists January 23, 2007

Posted by dfhuynh in research, semantic web.

I really want to write a nice post today, but I don’t know how to begin! This is simply too exciting to me! So, let’s start with what has just happened:

First, Johan Sundström made his Instant Google Spreadsheets Exhibit, which lets you turn any Google spreadsheet (with certain formatting requirements) into an exhibit just by pasting in its public feed URL. That gives you instant faceted browsing on a spreadsheet! With a bit more hacking, I’m sure he can let you switch to map and time line views, too! Think of editing data in the friendly UI of Google Spreadsheets and mashing up several spreadsheets into a faceted browsing UI with maps, time lines, histograms, …

Then, Michael Bergman converted his comprehensive list of Semantic Web tools to an exhibit and embedded it right inside his blog!

Let’s take 5 minutes to imagine the possibilities: any average blogger, on any day, who has a little bit of structured data in any domain, can easily enter the data into a Google spreadsheet and embed an exhibit in a blog post or a blog page. The data is published in a faceted browsing UI, on maps, time lines, histograms, etc. It can be copied off in various formats including RDF/XML, Semantic MediaWiki text, Tab-Separated Values, … It can be readily harvested by Piggy Bank, and re-published to a Semantic Bank

Now, we’re not just talking about the usual suspects of domains like photos and bookmarks. We’re talking about any arbitrary domain of data: the stamp collection your grandma left to you, the push puppet collection you collected for who knows why, … Yes, the real world data. The data that people care about.

I am hoping that we can create tools that let the average blogger easily publish structured data without ever having to coin URIs or design or pick ontologies…

Well, I don’t know how to end this post, either. But there it is: this is, in my humble opinion, a beginning of something great.

Thinking Inside the Box December 7, 2006

Posted by dfhuynh in research, semantic web.
add a comment

Too many people are thinking outside the box these days. I, I think inside the box.

In 2004, I put a web server inside the web browser (Piggy Bank). Now I put a database-backed 3-tier web application inside a single web page (Exhibit).

if Semantic Web is the medium, what is the message? May 20, 2006

Posted by dfhuynh in research, semantic web.

A colleague of mine is in the habit of speaking in URLs. Often for points he wishes to illustrate, he would dig out from the vast Web a URL and give us that, instead. I’ve jokingly noted that ours is turning into a URL-speaking society, but there is truth in that. The Web, from one point of perspective, is a rudimentary global language in which the URLs are the nouns. And everything else necessary for a language is missing.

Then comes the Semantic Web.

Designed to fill in the slots for everything that the Web is missing, the Semantic Web is a hot medium. It leaves little for the reader’s imagination. Indeed, it’s supposed to be so perfectly specified that machines can reliably act on it.

Web 2.0, on the other hand, is a cool medium that calls for participation, for social interaction, for interpretation and involvement. Tagging, blogging, etc. all require heavy participation and interpretation, which makes Web 2.0 a lot more congruent with our cool electric society today. Hence, Web 2.0 has easily taken off and Semantic Web is left in the lab cage.

Perhaps just like many great ideas before it, the Semantic Web is here at the wrong time, or at least cast in the wrong way to our particular society at its particular state.


So then, if Semantic Web is the medium, what is its message? The message that is understandable by the electric man, connected with cell phones, blogs, tags,…, who is becoming once again the tribal man as Marshall McLuhan predicted.

In many ways, the Semantic Web is like the printing technology (movable types). Whole books can be beautifully, precisely, and multiply produced by stringing together repeatable, uniform units that are carefully placed. Similarly, complex topics can now be described (supposedly) precisely by connecting together as graphs of URIs (supposedly) carefully coined. The similarity between print and Semantic Web is clear when one looks at how sophisticated Semantic Web UIs are rendered: views are recursively embedded one within another as the machine reaches out into the graph and pulls in more and more data. This gave rise to my obsession with the “fine granularity” that RDF affords when I started my research on the Haystack project. That is, just as typography is about repeatability, RDF is about recursability.

Just as print, according to McLuhan, turned the tribal man into the individualistic man, capable of entertaining thoughts outside communal gatherings, Semantic Web is an individualistic technology. It is about gathering information from external sources and then interpreting it in one’s own way for one’s own individual needs. That is, roll your own semantics on someone else’s data. This is why my Piggy Bank research has been so instrumental for understanding the Semantic Web for many, and why Simile‘s attempts at a screen-scraper ecology and Semantic Bank deployments have been less than successful.

Just like movable type made words more affordable, the Semantic Web will make data more affordable as data can be more easily re-used independent of its origin and original purpose. And just as printed words are mediated by paper and ink, Semantic Web data is necessarily mediated by the machine that gather, interpret, and combine disparate Semantic Web data into a coherent soup upon each individual’s demand.

That is all to say, the Semantic Web is like typography for this primitive, global language called the Web. And we are still searching for its typesetting machine.


When the Semantic Web starts to mature, the artists will move in and explore this new medium. Expect machine processible semantic puns. Expect graph arts.

Understanding Media May 19, 2006

Posted by dfhuynh in research, semantic web.

Four/five years ago I could hardly make it past the first chapter of Understanding Media by Marshall McLuhan, from which came the infamous sentence, "the medium is the message." I have recently returned to it and I'm now half-way through, enjoying every page. And although I'm not sure if I can re-verbalize his ideas, I do find the book very illuminating. I'd recommend it even more than I'd recommend Edward Tufte's books.

This is my effort to try to understand this new medium called the Semantic Web on which I'm doing my research.

Core to McLuhan's book, as how I understand it, is the idea that technologies shape the way humans fundamentally think and behave, regardless of how the technologies are used, that each technology bears some intrinsic characteristics independent of its use. And that humans, embedded within the technologies, cannot but think and act in some particular ways. This is a generalization of the Sapir-Whorfian hypothesis which posits that one's thoughts are moulded by whichever native language that one speaks.

The most primitive tribes of Australia and Africa, like the Eskimos of today, have not yet reached finger-counting, nor do they have numbers in series. Instead they have a binary system of independent numbers for one and two, with composite numbers up to six. After six, they perceive only "heap." Lacking the sense of series, they will scarcely notice when two pins have been removed from a row of seven. They become aware at once, however, if one pin is missing. [p.111]

The computer science researchers these days speak a few certain technical languages and are embedded within a few certain technologies. For example, some speak the "relational database" language, some speak the "Emac buffer" language, some speak the "Perl" language, … And a lot of the Semantic Web researchers speak some sort of "graph" language, although many are still stuttering and think in the "relational database" language.

For those who speak the "relational database" language, it is next to impossible to comprehend a world of graphs wherein the cost of creating relationships is zero. So they don't speak of "relationships" although ironically they deal with relational databases. They only speak of foreign keys and tables. They even have tables that contain only 2 foreign keys… If you ask me, it's another case of premature optimizations.

The Semantic Web folks are further along. They have adopted the "graph" language but then become greatly incapable of talking to the rest of the world who don't speak graphs. In vain, they take graphs and splash them on the screen, literally, creating visualizations full of nodes and arrows that are incomprehensible to everyone else.

But upon closer inspection, most of these graphs that they create are shallow, and some are just trees. It turns out, I suspect, that the Semantic Web people are simply obsessed with graphs that they try to turn everything into graphs even if they can only find shallow, tabular data. And so they are back to square one, busy exploring tools for visualizing and dealing with tabular data in graph's clothing without even realizing it.

It is amusing, puzzling, and frustrating to watch the great minds, many way "smarter" than I am, getting stuck in their own mindsets.

Once men have adopted the visual dynamic of the phonetic alphabet, they begin to lose the tribal man's obsession with cosmic order and ritual as recurrent in the physical organs and their social extension. Indifference to the cosmic, however, fosters intense concentration on minute segments and specialist tasks, which is the unique strength of Western man. For the specialist is one who never makes small mistakes while moving toward the grand fallacy. [p.124]

It is so true it is terrifying. After all, what mindset am I getting stuck in?! I will never know.