Re: sentence in NLP to ontology query?

Dear Xavier, all,

 here follow a few pointers on the issue of querying a knowledge
base through natural language:

There has been a lot of early work on the issue (typically referred to
as Natural Language Interfaces) since the late 70s, especially in
the Computational Linguistics community. These early natural language
interfaces (NLIs) constituted a front-end to databases (even before
the relational model was actually born). The problem was certainly
that these interfaces were not portable. A very good overview of
this early research can be found in the following two surveys containing
a lot of pointers to related work. In particular, you
will also find there some pointers to menu-based frontends to
databases.

@Article{nli_intro,
    author = "I. Androutsopoulos and G.D. Ritchie and P. Thanisch",
    title = "Natural Language Interfaces to Databases--An introduction",
    journal = "Journal of Language Engineering",
    volume = "1",
    number = "1",
    pages = "29--81",
    year = "1995"
}

See here: http://citeseer.ist.psu.edu/androutsopoulos95natural.html

and also:

@Article{nli_intro2,
    author = "A. Copestake and K. Sparck Jones",
    title = "Natural Language Interfaces to Databases",
    journal = "Knowledge Engineering Review",
    year = "1989",
    note = "Special Issue in the Applications of Natural Language
           Processing Techniques"
}

See here: http://www.cl.cam.ac.uk/~aac10/papers/ker.ps.gz


The main issues involved in building natural language interfaces are (i)
portability (to other domains) (ii) reliability, (iii) coverage and
(iv) intuitiveness. 

Recent work addressing the reliability/coverage aspects can be found
for example in:

http://www.cs.washington.edu/homes/amp/iui03.pdf

Old work addressing the portability can be found in:

@Article{team,
  author = 	 {B.J. Grosz and D.E. Appelt and P.A. Martin and F.C.N.
Pereira},
  title = 	 {TEAM: An Experiment in the Design of Transportable Natural
                  Language Interfaces},
  journal = 	 {Artificial Intelligence},
  year = 	 {1987},
  volume = 	 {32},
  pages = 	 {173-243}
}


@InCollection{rendezvous,
  author = 	 {E.F. Codd},
  title = 	 {Seven Steps to {RENDEZVOUS} with the Casual User},
  booktitle = 	 {Data Base Management},
  publisher =    {North-Holland publishers},
  year = 	 {1974},
  editor = 	 {J. Kimbie and K. Koffeman}
 }

@Article{ask,
  author = 	 {B.H. Thompson and F.B. Thompson},
  title = 	 {{ASK} is transportable in Half a Dozen Ways.},
  journal = 	 {ACM Transactions on Office Information Systems},
  year = 	 {1985},
  volume = 	 {3},
  number = 	 {2},
  pages = 	 {185--203},
}

A recent system using Semantic Web ontologies as backend is Aqualog:

http://kmi.open.ac.uk/projects/akt/aqualog/

Recent approaches using restricted language can be found here:

http://www.ifi.unizh.ch/ddis/staff/goehring/btw/files/BernsteinEtAl_ISWC2005.pdf

http://www.ifi.unizh.ch/attempto/


Finally, here are some specific workshops on the topic in the last two
years:

@Proceedings{qa_restricted_domains_04,
  title = 	 {ACL 2004 Workshop on Question Answering in Restricted
Domains },
  year = 	 {2004},
  editor = 	 {Diego Molla},
}

@Proceedings{qa_restricted_domains_05,
  title = 	 {AAAI 2005 Workshop on Question Answering in Restricted
Domains },
  year = 	 {2005},
  editor = 	 {Diego Molla},
}

@Proceedings{pragmatics_qa,
  title = 	 {HLT/NAACL Workshop on Pragmatics of Question Answering},
  year = 	 {2004},
  editor = 	 {???},
}

@Proceedings{taln_qa,
  title = 	 {TALN 2004 Question-Answering Workshop},
  year = 	 {2004},
  editor = 	 {???},
}

Last not least, some forthcoming journal issues on the topic:

@Proceedings{qa_journal1,
  title = 	 {Questions and Answers: Theoretical and Applied
Perspective},
  year = 	 {2005},
  booktitle =    {Special Issue of the Journal of Applied Logic},
  editor = 	 {R. Bernardi and B. Webber},
  note = 	 {forthcoming},
}

@Proceedings{qa_journal2,
  title = 	 {Question Answering in Restricted Domains},
  year = 	 {2005},
  booktitle =    {Special Issue of Computational Linguistics},
  editor = 	 {D. Molla and J.L Vicedo},
  note = 	 {forthcoming},
}

Of course, the above list is far from complete, but it's a start...

Best,

  Philipp.

On Sun, 2005-10-23 at 20:37 -0700, Richard H. McCullough wrote:
> I can give you three data points from my own personal experience.
> 
> 1. I have abandoned the quest for natural language, and invented
> my own knowledge representation language, MKR, which is 
> sufficiently English-like that non-experts can use it as an interface.
> Queries are expressed in MKR, and executed by my MKE program.
> My MKE program can translate MKR to/from OWL.
> 
> 2. OpenCyc intends to have an English interface to their KB, but
> they haven't reached their goal as yet.  They are currently working
> on an OWL interface to the KB.  
> OpenCyc has a primitive CycL-to-English generator operational.
> They have no English-to-CycL translator.  
> They do have a "dictionary assistant" which uses WordNet 
> to guide the user in translating from English to the corresponding 
> CycL concept.  (I haven't used the dictionary assistant yet, 
> that's on my to-do list.)  OpenCyc also has a KB browser which 
> searches for matching CycL concepts, which is very useful.  
> In this case, it is the human who is translating from English to CycL.
> 
> 3. There are numerous researchers working on KB systems
> which use English as the knowledge representation language.
> In this scenario, there is no translation required.
> 
> Dick McCullough
> knowledge := man do identify od existent done;
> knowledge haspart proposition list;
> http://rhm.cdepot.net/
> ----- Original Message ----- 
> From: "Xavier Noria" <fxn@hashref.com>
> To: <semantic-web@w3.org>
> Sent: Sunday, October 23, 2005 2:17 PM
> Subject: sentence in NLP to ontology query?
> 
> 
> > 
> > I have seen in several places those questions in natural language  
> > that are said to be addressable with, say, the expressiveness and  
> > logic properties of OWL DL. For instance the OWL Guide starts with
> > 
> >     Tell me what wines I should buy to serve with each course
> >     of the following menu. And, by the way, I don't like Sauternes.
> > 
> > However, I have never seen an algorithm mentioned, nor even a  
> > description of some approach, for how the translation from natural  
> > language to some query language can be automated. Whether some  
> > constraints have to be assumed in the natural language interface,  
> > whether it is even possible or not such a translation, etc.
> > 
> > Are there any pointers that can help me get the picture? Which is the  
> > state of the art? Or do people always mean the translation as of  
> > today has to be done by a person?
> > 
> > -- fxn
> > 
> > PS: I think the OWL Guide may say at the end something about this,  
> > because it starts that way and then it does not make clear it is only  
> > going to present a formal and thus partial solution to the problem.  
> > You end up wondering about the size of the hole between what you have  
> > actually read and the solution to the original opening problem.
> > 
> > 
> >
> 
>

Received on Monday, 24 October 2005 12:54:37 UTC