Category: Semantic Web

Concept Maps: Training Children to Build Ontologies?

May 31st, 2005 — 11:51am

Con­cept maps popped onto the radar last week when an arti­cle in Wired high­lighted a con­cept map­ping tool called Cmap. Cmap is one of a vari­ety of con­cept map­ping tools that’s in use in schools and other edu­ca­tional set­tings to teach chil­dren to model the struc­ture and rela­tion­ships con­nect­ing — well — con­cepts.
The root idea of using con­cept map­ping in edu­ca­tional set­tings is to move away from sta­tic mod­els of knowl­edge, and toward dynamic mod­els of rela­tion­ships between con­cepts that allow new kinds of rea­son­ing, under­stand­ing, and knowl­edge. That sounds a lot like the pur­pose of OWL.
It might be a stretch to say that by advo­cat­ing con­cept maps, schools are in fact train­ing kids to cre­ate ontolo­gies as a basic learn­ing and teach­ing method, and a vehi­cle for com­mu­ni­cat­ing com­plex ideas — but it’s a very inter­est­ing stretch all the same. As Infor­ma­tion Archi­tects, we’re famil­iar with the ways that struc­tured visu­al­iza­tions of inter­con­nected things — pages, top­ics, func­tions, etc. — com­mu­ni­cate com­plex notions quickly and more effec­tively than words. But most of the rest of the world doesn’t think and com­mu­ni­cate this way — or at least isn’t con­sciously aware that it does.
It seems rea­son­able that kids who learn to think in terms of con­cept maps from an early age might start using them to directly com­mu­ni­cate their under­stand­ings of all kinds of things through­out life. It might be a great way to com­mu­ni­cate the com­plex thoughts and ideas at play when answer­ing a sim­ple ques­tion like “What do you think about the war in Iraq?“
Author Nancy Kress explores this excact idea in the sci­ence fic­tion novel ‘Beg­gars In Spain’, call­ing the con­struc­tions “thought strings”. In Kress’ book, thought strings are the pre­ferred method of com­munca­tion for extremely intel­li­gent genet­i­cally engi­neered chil­dren, who have in effect moved to realms of cog­ni­tive com­plex­ity that exceed the struc­tural capac­ity of ordi­nary lan­guages. As Kress describes them, the den­sity and mul­ti­di­men­sional nature of thought strings makes it much eas­ier to share nuanced under­stand­ings of extremely com­plex domains, ideas, and sit­u­a­tions in a com­pact way.
I’ve only read the first novel in the tril­ogy, so I can’t speak to how Kress devel­ops the idea of thought strings, but there’s a clear con­nec­tion between the con­struct she defines and the con­cept map as laid out by Novak, who says, “it is best to con­struct con­cept maps with ref­er­ence to some par­tic­u­lar ques­tion we seek to answer or some sit­u­a­tion or event that we are try­ing to under­stand”.
Excerpts from the Wired arti­cle:
“Con­cept maps can be used to assess stu­dent knowl­edge, encour­age think­ing and prob­lem solv­ing instead of rote learn­ing, orga­nize infor­ma­tion for writ­ing projects and help teach­ers write new cur­ric­ula. “
“We need to move edu­ca­tion from a mem­o­riz­ing sys­tem and repet­i­tive sys­tem to a dynamic sys­tem,” said Gas­par Tarte, who is spear­head­ing edu­ca­tion reform in Panama as the country’s sec­re­tary of gov­ern­men­tal inno­va­tion.“
“We would like to use tools and a method­ol­ogy that helps chil­dren con­struct knowl­edge,” Tarte said. “Con­cept maps was the best tool that we found.”

Comments Off | Modeling, Semantic Web

Semantic Ambiguity Strikes Your Local Pub

May 16th, 2005 — 4:31pm

Thurs­day night I was at Casablanca in Har­vard Square for an infor­ma­tion archi­tec­ture meet and greet after Lou’s Enter­prise IA sem­i­nar. I ordered a Wolver’s. It was dim and noisy, so after shout­ing three times and point­ing, I ended up with a Wolaver’s
Not a sur­prise, right? My first thought was “What’s in my glass?” My sec­ond thought — I was sur­rounded by infor­ma­tion archi­tects — was about the seman­tic angle on the sit­u­a­tion. It seems like a fair mis­take to make in a loud and crowded bar. But as some­one who works there, he should know the envi­ron­men­tal con­text, the ways it affects fun­da­men­tal tasks like talk­ing and answer­ing ques­tions, and about any alter­na­tives to what he thought I said that are close enough to be eas­ily mis­taken. Before I get too far, I’ll point out that I liked the mis­take enough to order another.
Set­ting aside for a moment the notion of a seman­ti­cally adept agent sys­tem that mon­i­tors inter­ac­tions between bar­tenders and patrons to pre­vent mis­takes like this, let’s look at some­thing more likely, such as how does Google fair with this sit­u­a­tion? Some post-socialization research shows that as far as Google is con­cerned, all roads do in fact lead to Wolaver’s. Even when Google’s results list begins with a link to a page on Wolver’s Ale from the orig­i­nat­ing brew­ery, it still sug­gests that you might want ‘wolaver’s ale’. Maybe this explains the bartender’s mis­take.
Here’s the break­down: Google US sug­gests “wolaver’s ale” when you search for “wolvers ale” and “wolver’s ale”, but not the other way around. When you search for “Wolavers”, Google sug­gests the cor­rectly punc­tu­ated “Wolaver’s”. You can get to the Amer­i­can ale, but not the British.
More sur­pris­ing, it’s the same from Google UK, when search­ing only British pages. (Some­one tell me how pages become part of the UK? Maybe when they’re sent off to full-time board­ing school?)
Google’s insis­tence on tak­ing me from wher­ever I start to “Wolaver’s Ale” comes from more than sim­ple Amer­i­can brew chau­vanism. This is what hap­pens when the wrong fac­tors drive deci­sions about the mean­ings of things; it’s these basic deci­sions about seman­tics that deter­mine whether or not a thing cor­rectly meet the needs of the peo­ple look­ing for answers to a ques­tion.
You might say seman­tic mis­align­ment (or what­ever we choose to call this con­di­tion) is fine, since Google’s busi­ness is aimed at doing some­thing else, but I can’t imag­ine that busi­ness leaderhsip and staff at Wolver’s would be too happy to see Google direct­ing traf­fic away from them by sug­gest­ing that peo­ple didn’t want to find them in the first place. Nei­ther Wolver’s nor Wolavers seems to have Google ads run­ning for their names, but what if they did? By now we’re all fami­lar with the fact that googling ‘mis­er­able fail­ure’ returns a link to the White House web site. This reflects a pop­u­larly defined asso­ci­a­tion rich in cul­tural sig­nif­i­cance, but that isn’t going to sat­isfy a pay­ing cus­tomer who is los­ing busi­ness because a seman­ti­cally unaware sys­tem works against them.
This a good exam­ple of a sit­u­a­tion in which intel­li­gent dis­am­bigua­tion based on rela­tion­ships and infer­enc­ing within a defined con­text has direct busi­ness ram­i­fi­ca­tions.
Here’s a pre­view of the full size table that shows the results of check­ing some vari­ants of wolvers / wolavers:

Comments Off | Semantic Web

NY Times Mistake Shows Utility of Semantic Framework

April 25th, 2005 — 10:09pm

Read­ing the online edi­tion of the New York Times just before leav­ing work this after­noon, I came across an ironic mis­take that shows the util­ity of a well devel­oped seman­tic frame­work that mod­els the terms and rela­tion­ships in defin­gin dif­fer­ent edi­to­r­ial con­texts. In an arti­cle dis­cussing the Matrix Online mul­ti­player game, text iden­ti­fy­ing the movie char­ac­ter the Ora­cle mis­tak­enly linked to a busi­ness pro­file page on the com­pany of the same name. In keep­ing with the movie’s sin­is­ter depic­tions of tech­nol­ogy as a tool for cre­at­ing decep­tive medi­ated real­i­ties, by the time I’d dri­ven home and made moji­tos for my vis­it­ing in-laws, the mis­take was cor­rected…
Ironic humor aside, it’s unlikely that NYTimes Dig­i­tal edi­tors intended to con­fuse a movie char­ac­ter with a giant soft­ware com­pany. It’s pos­si­ble that the NYTimes Dig­i­tal pub­lish­ing plat­form uses some form of seman­tic frame­work to over­see auto­mated link­ing of terms that exist in one or more defined ontolo­gies, in which case this mis­take implies some form of mis-categorization at the arti­cle level,invokgin the wrong ontol­ogy. Or per­haps this is an exam­ple of an instance where a name in the real world exists simul­ta­ne­ously in two very dif­fer­ent con­texts, and there is no seman­tic rule to gov­ern how the sys­tem han­dles rec­on­cil­i­a­tion of con­flicts or invo­ca­tion of man­ual inter­ven­tion in cases when life refuses to fit neatly into a set of ontolo­gies. That’s a design fail­ure in the gov­er­nance com­po­nents of the seman­tic frame­work itself.
It’s more likely that the pub­lish­ing plat­form auto­mat­i­cally searches for com­pany names in arti­cles due for pub­li­ca­tion, and then cre­ates links to the cor­re­spond­ing pro­file infor­ma­tion page with­out ref­er­ence to a seman­tic frame­work that employs con­tex­tual mod­els to dis­crim­i­nate between ambigu­ous or con­flict­ing term usage. For a major con­tent cre­ator and dis­trib­u­tor like the NY Times, that’s a strate­gic over­sight.
In this screen cap­ture, you can see the first ver­sion of the arti­cle text, with the link to the Ora­cle page clearly vis­i­ble:

The new ver­sion, with­out the mis­taken link, is vis­i­ble in this screen cap­ture:
New Ver­sion:

Comments Off | Semantic Web

mSpace Online Demo

February 20th, 2005 — 2:48pm

There’s an mSpace demo online.

Comments Off | Modeling, Semantic Web, User Experience (UX)

Two Surveys of Ontology / Taxonomy / Thesaurus Editors

February 18th, 2005 — 2:46pm

While research­ing and eval­u­at­ing user inter­faces and man­age­ment tools for seman­tic struc­tures — ontolo­gies, tax­onomies, the­sauri, etc — I’ve come across or been directed to two good sur­veys of tools.
The first, cour­tesy of HP Labs and the SIMILE project is Review of exist­ing tools for work­ing with schemas, meta­data, and the­sauri. Thanks to Will Evans for point­ing this out.
The sec­ond is a com­pre­hen­sive review of nearly 100 ontol­ogy edi­tors, or appli­ca­tions offer­ing ontol­ogy edit­ing capa­bil­i­ties, put together by Michael Denny at You can read the full arti­cle Ontol­ogy Build­ing: A Sur­vey of Edit­ing Tools, or go directly to the Sum­mary Table of Sur­vey Results.
The orig­i­nal date for this is 2002 — it was updated July of 2004.

Comments Off | Modeling, Semantic Web, User Experience (UX)

mSpace: A New (Usable?) Semantic Web Interface

February 18th, 2005 — 10:56am

mSpace is a new frame­work — includ­ing user inter­face — for inter­act­ing with seman­ti­cally struc­tured infor­ma­tion that appeared on Slash­dot this morn­ing.
Accord­ing to the sup­port­ing lit­er­a­ture, mSpace han­dles both onto­log­i­cally struc­tured data, and RDF based infor­ma­tion that is not mod­elled with ontolo­gies.
What is poten­tially most valu­able about the mSpace frame­work is a use­ful, usable inter­face for both nav­i­gat­ing / explor­ing RDF-based infor­ma­tion spaces, and edit­ing them.
From the mSpace source­forge site:
“mSpace is an inter­ac­tion model designed to allow a user to nav­i­gate in a mean­ing­ful man­ner the multi-dimensional space that an ontol­ogy can pro­vide. mSpace offers poten­tially use­ful slices through this space by selec­tion of onto­log­i­cal cat­e­gories.
mSpace is fully gen­er­alised and as such, with a lit­tle def­i­n­i­tion, can be used to explore any knowl­edge base (with­out the require­ment of ontolo­gies!).
Please see for more infor­ma­tion.“
From the abstract of the Tech­ni­cal report, titled mSpace: explor­ing the Seman­tic Web
“Infor­ma­tion on the web is tra­di­tion­ally accessed through key­word search­ing. This method is pow­er­ful in the hands of a user that is expe­ri­enced in the domain they wish to acquire knowl­edge within. Domain explo­ration is a more dif­fi­cult task in the cur­rent envi­ron­ment for a user who does not pre­cisely under­stand the infor­ma­tion they are seek­ing. Seman­tic Web tech­nolo­gies can be used to rep­re­sent a com­plex infor­ma­tion space, allow­ing the explo­ration of data through more pow­er­ful meth­ods than text search. Ontolo­gies and RDF data can be used to rep­re­sent rich domains, but can have a high bar­rier to entry in terms of appli­ca­tion or data cre­ation cost.
The mSpace inter­ac­tion model describes a method of eas­ily rep­re­sent­ing mean­ing­ful slices through these mul­ti­di­men­sional spaces. This paper describes the design and cre­ation of a sys­tem that imple­ments the mSpace inter­ac­tion model in a fash­ion that allows it to be applied across almost any set of RDF data with min­i­mal recon­fig­u­ra­tion. The sys­tem has no require­ment for onto­log­i­cal sup­port, but can make use of it if avail­able. This allows the visu­al­i­sa­tion of exist­ing non-semantic data with min­i­mal cost, with­out sac­ri­fic­ing the abil­ity to utilise the power that semantically-enabled data can provide.”

Comments Off | Semantic Web, User Experience (UX)

Public RDF Data Sets at

February 8th, 2005 — 9:56am offers a great col­lec­tion of RDF data sets and ser­vices that gen­er­ate RDF.

Comments Off | Semantic Web

Tim Bray and the RDF Challenge: Poor Tools Are A Barrier For The Semantic Web

February 7th, 2005 — 4:56pm

In the lat­est issue of ACMQueue, Tim Bray is inter­viewed about his career path and early involve­ment with the SGML and XML stan­dards. While recount­ing, Bray makes four points about the slow pace of adop­tion for RDF, and reit­er­ates his con­vic­tion that the cur­rent qual­ity of RDF-based tools is an obsta­cle to their adop­tion and the suc­cess of the Seman­tic Web.
Here are Bray’s points, with some com­men­tary based on recent expe­ri­ences with RDF and OWL based ontol­ogy man­age­ment tools.
1. Moti­vat­ing peo­ple to pro­vide meta­data is dif­fi­cult. Bray says, “If there’s one thing we’ve learned, it’s that there’s no such thing as cheap meta-data.“
This is plainly a prob­lem in spaces much beyond RDF. I hold the con­cept and the label meta-data itself partly respon­si­ble, since the term meta-data explic­itly sep­a­rates the descriptive/referential infor­ma­tion from the idea of the data itself. I wager that user adop­tion of meta-data tools and processes will increase as soon as we stop dis­so­ci­at­ing a com­plete pack­age into two dis­tinct things, with dif­fer­ent implied lev­els of effort and value. I’m not sure what a uni­fied label for the base level unit con­struct made of meta-data and source data would be (an asset maybe?), but the implied deval­u­a­tion of meta-data as an optional or sup­ple­men­tal ele­ment means that the time and effort demands of accu­rate and com­pre­hen­sive tag­ging seem oner­ous to many users and busi­nesses. Thus the pro­lif­er­a­tion of auto­mated tax­on­omy and cat­e­go­riza­tion gen­er­a­tion tools…
2. Infer­ence based pro­cess­ing is inef­fec­tive. Bray says, “Infer­ring meta-data doesn’t work… Infer­ring meta-data by nat­ural lan­guage pro­cess­ing has always been expen­sive and flaky with a poor return on invest­ment.“
I think this isn’t spe­cific enough to agree with with­out qual­i­fi­ca­tion. How­ever, I have seen analy­sis of a num­ber of infer­renc­ing sys­tems, and they tend to be slow, espe­cially when pro­cess­ing and updat­ing large RDF graphs. I’m not a sys­tems archi­tect or an engi­neer, but it does seem that none of the var­i­ous solu­tions now avail­able directly solves the prob­lem of allow­ing rapid, real-time infer­renc­ing. This is an issue with struc­tures that change fre­quently, or dur­ing high-intensity peri­ods of the ontol­ogy life-cycle, such as ini­tial build and edi­to­r­ial review.
3. Bray says, “To this day, I remain fairly uncon­vinced of the core Seman­tic Web propo­si­tion. I own the domain name I’ve offered the world the chal­lenge, which is that for any­body who can build an actual RDF-based appli­ca­tion that I want to use more than once or twice a week, I’ll give them I announced that in May 2003, and noth­ing has come close.“
Again, I think this needs some clar­i­fi­ca­tion, but it brings out a seri­ous poten­tial bar­rier to the suc­cess of RDF and the Seman­tic Web by show­cas­ing the poor qual­ity of exist­ing tools as a direct neg­a­tive influ­encer on user sat­is­fac­tion. I’ve heard this from users work­ing with both com­mer­cial and home-built seman­tic struc­ture man­age­ment tools, and at all lev­els of usage from core to occa­sional.
To this I would add the idea that RDF was meant for inter­pre­ta­tion by machines not peo­ple, and as a con­se­quence the basic user expe­ri­ence par­a­digms for dis­play­ing and manip­u­lat­ing large RDF graphs and other seman­tic con­structs remain unre­solved. Mozilla and Netscape did won­ders to make the WWW appar­ent in a vis­ceral and tan­gi­ble fash­ion; I sus­pect RDF may need the same to really take off and enter the realm of the less-than-abstruse.
4. RDF was not intended to be a Knowl­edge Rep­re­sen­ta­tion lan­guage. Bray says, “My orig­i­nal ver­sion of RDF was as a general-purpose meta-data inter­change facil­ity. I hadn’t seen that it was going to be the basis for a general-purpose KR ver­sion of the world.“
This sounds a bit like a warn­ing, or at least a strong admo­ni­tion against reach­ing too far. OWL and vari­ants are new (rel­a­tively), so it’s too early to tell if Bray is right about the scope and ambi­tion of the Seman­tic Web effort being too great. But it does point out that the con­text of the stan­dard bears heav­ily on its even­tual func­tional achieve­ment when put into effect. If RDF was never meant to bear its cur­rent load, then it’s not a sur­prise that an effec­tive suite of RDF tools remains unavailable.

Comments Off | Semantic Web

Semantic Mapping, Ontologies, and XML Standards

May 3rd, 2004 — 10:23am

Here’s a some snip­pets from an arti­cle in the Web Ser­vices Jour­nal that nicely explains some of the busi­ness ben­e­fits of a services-based archi­tec­ture that uses ontolo­gies to inte­grate dis­parate appli­ca­tions and knowl­edge spaces.
Note that XML / RDF / OWL — all from the W3C — together only make up part of the story on new tools for how mak­ing it easy for sys­tems (and users, and busi­nesses…) to under­stand and work with com­pli­cated infor­ma­tion spaces and rela­tion­ships. There’s also Topic Maps, which do a very good job of visu­ally map­ping rela­tion­ships that peo­ple and sys­tems can under­stand.
Seman­tic Map­ping, Ontolo­gies, and XML Stan­dards
The key to man­ag­ing com­plex­ity in appli­ca­tion inte­gra­tion projects
Another impor­tant notion of ontolo­gies is entity cor­re­spon­dence. Ontolo­gies that are lever­aged in more of a B2B envi­ron­ment must lever­age data that is scat­tered across very dif­fer­ent infor­ma­tion sys­tems, and infor­ma­tion that resides in many sep­a­rate domains. Ontolo­gies in this sce­nario pro­vide a great deal of value because we can join infor­ma­tion together, such as prod­uct infor­ma­tion mapped to on-time deliv­ery his­tory mapped to cus­tomer com­plaints and com­pli­ments. This estab­lishes entity cor­re­spon­dence.
So, how do you imple­ment ontolo­gies in your appli­ca­tion inte­gra­tion prob­lem domain? In essence, some tech­nol­ogy — either an inte­gra­tion bro­ker or appli­ca­tions server, for instance — needs to act as an ontol­ogy server and/or map­ping server.
An ontol­ogy server houses the ontolo­gies that are cre­ated to ser­vice the appli­ca­tion inte­gra­tion prob­lem domain. There are three types of ontolo­gies stored: shared, resource, and appli­ca­tion. Shared ontolo­gies are made up of def­i­n­i­tions of gen­eral terms that are com­mon across and between enter­prises. Resource ontolo­gies are made up of def­i­n­i­tions of terms used by a spe­cific resource. Appli­ca­tion ontolo­gies are native to par­tic­u­lar appli­ca­tions, such as an inven­tory appli­ca­tion. Map­ping servers store the map­pings between ontolo­gies (stored in the ontol­ogy server). The map­ping server also stores con­ver­sion func­tions, which account for the dif­fer­ences between schemas native to remote source and tar­get sys­tems. Map­pings are spec­i­fied using a declar­a­tive syn­tax that pro­vides reuse.
RDF uses XML to define a foun­da­tion for pro­cess­ing meta­data and to pro­vide a stan­dard meta­data infra­struc­ture for both the Web and the enter­prise. The dif­fer­ence between the two is that XML is used to trans­port data using a com­mon for­mat, while RDF is lay­ered on top of XML defin­ing a broad cat­e­gory of data. When the XML data is declared to be of the RDF for­mat, appli­ca­tions are then able to under­stand the data with­out under­stand­ing who sent it.

Comments Off | Semantic Web

Back to top