[DL] Mapping words or utterances (in natural language) to ontology concepts

Sebastien Ferre Sebastien.Ferre at irisa.fr
Mon Sep 9 18:31:28 CEST 2013


as your first motivation is not reasoning but mapping from
natural language to a formal language, you might consider
SPARQL as a target language. SPARQL 1.1 is very expressive
and allows for the expression of factual statements (as updates),
and all sorts of queries using expressions, aggregations,
Boolean operators, etc.

I work on a translator from controlled English (SQUALL) to SPARQL 1.1.
Web forms for translation and direct querying of SPARQL endpoints
are available from this page:


It might give you some idea of what is possible. It's only
for controlled English and the online version uses no lexicon
(qnames must be used for words), but I think it covers a pretty
large fragment of NL. A future version will allow to improve
naturalness by feeding in a lexicon (e.g., for DBpedia).

Your first example can be turned into a SQUALL question:

 > Which person has at least 2 son-s and at most 5 daughter-s ?
(assuming 'person' is a class, and 'son' and 'daughter' are properties)

Your second example, assuming a small extension not yet available
online, can be expressed as:

 > Which person has the number of son-s / the number of child-s greater 
than 1 / 3 ?
(assuming 'child' is a super-property of 'son' and 'daughter')

Using an ad-hoc lexicon would make it easy to translate the nouns
'son' and 'daughter' as a combination of the property 'child', and
the classes 'male' and 'female', like in your example.


On 13/08/13 13:45, Olivier Austina wrote:
> Hi,
> My question is about mapping a word or an utterance (in natural
> language) to ontology for Natural Language Interface to DataBase
> (NLIDB). The database can be an RDF graph or traditional database.
> The problem is that the lexicon derived from the ontology is very
> limited to cover how the natural language can express the concepts of
> the ontology. It is possible to enriched labels of the ontology with
> synonymy or hypernymy/hyponymy but it is still insufficient to cover the
> way natural language can refer to ontology concepts.
> The first motivation is not reasoning but be able to map natural
> language to ontology. I think Description Logic is suitable for this
> task by providing  a way a word or an  utterance can be describe
> according a given ontology but I don't know how far it is possible to go
> with DL (strength and limitation of DL for this kind of task).
> This an example taken from the DL handbook 2003 (p. 93) :
> for a person having at least 2 sons and at most 5 daughters.
> Somtimes, it is possible to have an expression or function instead of
> integer like 2 or 5 in this example to quantify roles or concepts. For
> example : A person with more than a third of her children are males. The
> expression or th function can be provided by a formal query language
> such as SQL or SPARQL or a user defined functions.
> My questions are the following:
> Can DL support an arbitrary expression, function?
> What can be expressed using DL and cannot be expressed (the motivation s
> not reasoning but mapping natural language to ontology concept).
> Any suggestion is welcome. Thanks.
> Regards
> Olivier
> ---
> **  You received this mail via the description logic mailing list; for more  **
> **  information, visit the description logic homepage at http://dl.kr.org/.  **

More information about the dl mailing list