[Attempto] the case of mixing RDF vocabularies : a FOAF example

[Norbert E. Fuchs]

Jean-Marc & Kaarel

I read your email exchange and I admire your efforts to massage an ACE text into a form that achieves a desired effect in OWL.

However, I was struck by

On 13 Jun 2012, at 13:59 , Jean-Marc Vanel wrote:

> Anyway I recognize that it's not very natural language-like .

and would like to recall that ACE was developed for people who need to use formal methods – like OWL – but are not familiar with them, or even do not want to be bothered by them at all. One case in point are medical doctors with whom we worked in several projects.

In my view, Jean-Marc's final sentence

(1) If P is a person that has-homepage something and has-current-project something then P is an n:interesting-contact.

– though accepted by APE – does not conform to ACE's philosophy spelled out above. It looks like the result of an OWL -> ACE translation, and is not even acceptable English.

To conform to ACE's philosophy, Jean-Marc's sentence could for instance be

(2) Every person who has a homepage and a current project is an interesting contact.

There are several possibilities to get from (2) to OWL provided that you want to use APE and the DRS -> OWL translator. As Kaarel suggested – though for different reasons – you could translate (2) into (1) which shouldn't be too difficult. Or you could submit (2) to APE and then modify the resulting DRS. One modification that you discussed already is using an appropriate lexicon to generate a DRS suitable as as start point for further modifications. Other modifications could be performed on the DRS until you arrive at one that you can submit to the DRS -> OWL translator that of course could also be modified.

Your problem reminds me strongly of the translations that I do in my reasoner RACE to generate clauses from ACE texts: (1) submit the ACE texts to APE to generate a DRS, (2) translate the DRS into FOL using a variation of the standard DRS -> FOL translator that applies a lot of additional modifications, (3) translate FOL into clauses using a variation of the standard FOL clausifier that again applies many additional modifications. The additional modifications in steps (2) and (3) serve various purposes, for instance getting the correct semantics of the copula "to be" and the correct semantics of collective plurals, both needed to achieve the expected deductions. Currently, I am adding arithmetic, lists and sets to RACE, and may need to add further translations between steps (1) and (2) since the operations on lists and sets are not defined as operators, and need to be expressed in ACE.

Best regards.

   --- nef