Re: state of the art

Dear all

I have tried to answer Paola's questions below:
> 1. is there a relationship between NLReduce and CE (controlled English)
>   
No there is not. We did our very initial prototype called SWAT with 
Attempto Controlled English (ACE, see http://attempto.ifi.uzh.ch/site/), 
which is developed by Norbert Fuchs and others at the University of 
Zurich (UZH). While the SWAT worked and we could get some interesting 
results out of some user testing we decided that SWAT would be at an 
unfair disadvantage over the other tools in our experiment, as it 
imposes more structure on the user then NLP-Reduce or Querix, whilst not 
providing the support of Ginseng/GINO.

If you want to know more about the ACE-based SWAT look up our ISWC'05 paper:
A. Bernstein, E. Kaufmann, A. Göhring, C. Kiefer, "Querying Ontologies: 
A Controlled English Interface for End-users", /4th International 
Semantic Web Conference (ISWC 2005)/ <http://iswc2005.semanticweb.org/>, 
November 2005, p. 112-126.
which can be found at:
http://www.ifi.uzh.ch/ddis/staff/goehring/btw/files/BernsteinEtAl_ISWC2005d.pdf

NLP-Reduce is not assuming any structure whatsoever in the entry and 
"just" tries to do best effort with the query it gets. More about 
NLP-reduce can be found in:
E. Kaufmann, A. Bernstein, L. Fischer, "NLP-Reduce: A "naïve" but 
Domain-independent Natural Language Interface for Querying Ontologies", 
/4th European Semantic Web Conference (ESWC 2007)/ 
<http://kmi.open.ac.uk/events/eswc07/>, November 2007
http://www.ifi.uzh.ch/ddis/staff/goehring/btw/files/Kaufmann_NLP-Reduce_ESWC2007.pdf

Also, all the four tools presented in the talk are briefly discussed in 
the associated paper:
E. Kaufmann, A. Bernstein, "How Useful are Natural Language Interfaces 
to the Semantic Web for Casual End-users?", /6th International Semantic 
Web Conference (ISWC 2007)/ 
<http://iswc2007.semanticweb.org/main/default.asp>, 2007, p. 281-294
http://www.ifi.uzh.ch/ddis/staff/goehring/btw/files/Kaufmann_ISWC2007.pdf
The paper also has references to further literature about for the query 
interfaces.

> 2. assuming that these tools being discussed work with unstructured
> information (not published on the web in rdf/owl) - does that show
> that 'reasoning' can be done by the tool, not requiring the content to
> be expressed in owl/rdf?
>   
I am confused about this question. All four tools investigated operate 
directly on an OWL KB. The one used in the experiment can be found at:
http://www.ifi.uzh.ch/ddis/research/semweb/talking-to-the-semantic-web/owl-test-data/

> 2a. if so, can these tools be used to create ontologies, if they
> provided a functionality that allows users to add, say, axioms and or
> translate the NL process into some level of formalised logic?
>   
Yes. GINO/Ginseng, was actually used to do some very simple ontology 
editing tasks by end-users in our ISWC'06 paper.
A. Bernstein, E. Kaufmann, "GINO - A Guided Input Natural Language 
Ontology Editor", in proceedings of the 5th International Semantic Web 
Conference (ISWC 2006), Athens, Georgia, 2006, p. 144-157.
http://www.ifi.uzh.ch/ddis/staff/goehring/btw/files/BernsteinEtAl_Gino_ISWC_2006.pdf

> (I am referring to querix, for example, which allows users to select
> statements such as biggest city by population vs biggest by area - if
> expert users could interact with a tool, by browsing the web or
> selected knowledge bases? - can an ontology be inferred using
> the reasoning of an expert user if the tool is set up to capture the
> inferences and match them against existing ontologies/KBases?
We did not try it with Querix, but I do think that it would be possible. 
I know there are a bunch of folks doing exciting stuff on learning 
ontologies. Usually, they do it from text, but I do not see why you 
shouldn't be able to do so from observing user behavior. Actually, a 
very related question is one of the first questions that spawned machine 
learning to start with...
> 3.  should we carry out this study (or similar ones) with interfaces
> in different languages, and in different countries, and see if the
> results are comparable? (this would come under the scope of our
> current interest with the global perspective)
>   
The answer is, probably yes. But the tests are a lot of work!
Note, however, that even though all tests were carried out in English 
most subjects weren't English native speakers. This is further 
supporting our findings as they are probably even less comfortable to 
use English than their native language. Hence, even a less desirable 
natural language was preferred over a logic-based one.

We did ask the subjects about their prior knowledge/familiarity with 
English and this did not have any significant outcome on the results 
(see page 289) of the ISWC'07 paper.

I hope all of this helps

Cheers

Avi

Received on Thursday, 4 September 2008 15:34:01 UTC