-*- org -*-
This is dlvhex-dlplugin, a plugin for `dlvhex’.
This file includes some information regarding installation and needed tools and libraries in order to build the source package of dlvhex-dlplugin.
dlvhex-dlplugin depends on following libraries and programs:
http://www.kr.tuwien.ac.at/research/systems/dlvhex/
http://cppunit.sourceforge.net/
See dlvhex installation instructions.
Download the Racer/RacerPro binary and install it into your favourite location accourding to the installation instructions of Racer. Add the installation path to the PATH environment variable. dlvhex-dlplugin has been tested with RacerPro 1.8 and 1.9 and 1.9.2beta.
See the dlvhex installation instructions.
Popular Linux distributions provide raptor packages. On Debian/Ubuntu, just run
$ apt-get install libraptor-dev
If you want to install raptor from the sources, just follow the instructions of the INSTALL file that can be found in the distributions tarball. dlvhex-dlplugin needs at least libraptor 1.4.8 due to XML namespace parsing support.
See the dlvhex installation instructions.
To install dlvhex-dlplugin in to global dlvhex plugin directory, just run
$ ./configure
`make install’ will then copy the plugin to a directory where dlvhex can find it. E.g., usually to /usr/local/lib/dlvhex/plugins.
`configure –enable-userinstall’ will adjust the installation-path to `~/.dlvhex/plugins’, if the plugin should not be installed system-wide. Other configure-switches are shown with `configure –help’.
Other influental environment variables include:
Set this environment variable to the absolute path of dlvhex pkgconfig directory. Usually, this directory can be found in `$prefix/lib/pkgconfig’. E.g., run configure like
$ ./configure PKG_CONFIG_PATH=/usr/local/lib/pkgconfig
if dlvhex was installed with a prefix of /usr/local.
`configure’ searches for an executable called `RacerPro’ during its execution. If PATH is not setup properly, you can set the RACER variable to the absolute path of `RacerPro’:
$ ./configure RACER=/path/to/racerpro
Run `make check’ to build and execute the cppunit-based regression testsuite.
Run `make install’ to install dlvhex-dlplugin into a system-wide plugin directory provided by dlvhex.
The Description Logic Plugin interfaces HEX-programs with OWL ontologies by using a description logic reasoner. Currently, the DL reasoner Racer is supported. It provides five external atoms, roughly divided in
(i) querying atoms, i.e., external atoms which requests answers from the DL-reasoner to a given query, and (ii) the DL-KB consistency checking external atom, a boolean atom which decides whether augmenting a DL-KB with the given input parameters results in a consistent DL-KB.
The dl-plugin accepts the following command line arguments:
`–ontology=URI’: In cq-programs `(L,P)’, set the corresponding DL-KB `L’ to `URI’, where `URI’ is the location of an OWL Ontology in URI syntax. See the cq-program section for more information about cq-program usage.
`–dlopt=MOD[,MOD]*’: Setup particular optimization features according to the supplied list of modifiers `MOD’, which may be `-push’ for disabling push optimizations and `-cache’ for disabling the DL-Cache.
`–dldebug=LEVEL’: For debugging purposes, set `LEVEL’ accordingly to increase the verbosity of the log messages during query evaluation..
The atoms of the first category support query answering to concept (C) and role (R and DR) queries, conjunctive queries (CQ), and union of conjunctive queries (UCQ).
The querying external atoms have the form
(1) &dlT[KB,a,b,c,d,Q](X1,…,Xn),
where `T’ in `dlT’ denotes the type of the query and may be one of `C’, `R’, `DR’, `CQ’, and `UCQ’. `Q’ is then a query of type given by `T’, whereas `X1,…,Xn’ is an output tuple with arity matching the query answer.
The `&dlConsistent’ external atom has the form
(2) &dlConsistent[KB,a,b,c,d],
All external atoms share common input parameters. Their intended meaning is specified in the following Table.
—-—————————————————————-+
| KB | URI or file path of the OWL Ontology |
—-—————————————————————-+
| a | name of a binary predicate whose extension denotes addition to |
| a concept |
—-—————————————————————-+
| b | name of a binary predicate whose extension denotes addition to |
| the complement of a concept |
—-—————————————————————-+
| c | name of a ternary predicate whose extension denotes addition |
| to a role |
—-—————————————————————-+
| d | name of a ternary predicate whose extension denotes addition |
| to the complement of a role |
—-—————————————————————-+
A more detailed description of all external atoms, the proper syntax for `Q’, and the output list `X1,…,Xn’ is now subject of the following subsections.
Queries to concepts can be stated using the external atom
(3) &dlC[KB,a,b,c,d,Q](X),
where `Q’ is a concept name and `X’ is a term. If the external atom has a non-ground output, i.e., `X’ is a variable, then (3) retrieves all known members of concept `Q’. Otherwise, if `X’ is an individual, then (3) holds iff `X’ is an instance of concept Q.
Example 1: The following rule expresses a simple concept query.
wine(X) :- &dlC[“wine.rdf”,a,b,c,d,”Wine”](X).
Provided that `a’, `b’, `c’, and `d’ do not occur elsewhere in the HEX-program, this rule would do nothing else but putting all members of concept `Wine’ of the wine ontology `wine.rdf’ (see http://www.w3.org/TR/owl-guide/wine.rdf) into the extension of predicate wine.
The term `”Wine”’ is expressed in OWL and has thus an XML namespace attached to it. Since the concept `Wine’ uses the default namespace name `http://www.w3.org/TR/2003/PR-owl-guide-20031209/wine#’, we can simply refer to the concept of all wines as is. This would not be possible, if Wine is in the scope of a different namespace than the default namespace. Consider an ontology, where the concept of all wines is defined as follows:
<owl:Class rdf:ID=”vin:Wine”/>
Here, the term `”Wine”’ is bound to the XML namespace `vin’. To refer to the concept `vin:Wine’, which is short for the RDF/XML URI `http://www.w3.org/TR/2003/PR-owl-guide-20031209/wine#Wine’, we can use two different methods for doing so. First, we can use the fully expanded concept name `http://www.w3.org/TR/2003/PR-owl-guide-20031209/wine#Wine’ in the query part of the external atoms, or second, we introduce a namespace in the HEX-program itself. This can be accomplished by adding a namespace declaration to the program. Take, for instance, the program in Example 1. Adding the XML namespace vin to it results in the program
#namespace(vin,”http://www.w3.org/TR/2003/PR-owl-guide-20031209/wine#”). wine(X) :- &dlC[“wine.rdf”,a,b,c,d,”vin:Wine”](X).
Note that we can define an arbitrary namespace prefix for the namespace URI `http://www.w3.org/TR/2003/PR-owl-guide-20031209/wine#’.
Example 2: Now imagine we want to extend the wine concept by a new individual currently not known to be a wine. We use the concept augmentation mechanisms of the `&dlC’ atom.
wine(X) :- &dlC[“wine.rdf”,w,b,c,d,”Wine”](X). w(“Wine”,”Uhudler”).
In the program above, we add the individual `”Uhudler”’ to the concept `”Wine”’. The `&dlC’ atom expects a predicate name as second parameter; in this case w. This predicate must be binary and its first argument denotes the concept to be extended and the second the actual individual to be added to the concept.
This can be extended to even more flexible queries. In the next program, we supplement `”Uhudler”’ to `”Wine”’ and all italian redwines to the concept `”RedWine”’.
wine(X) :- &dlC[“wine.rdf”,w,b,c,d,”Wine”](X). w(“Wine”,”Uhudler”). w(“RedWine”,X) :- redwine(X), grows(X, italy).
Example 3: Similarly, we can increase the ABox of an ontology by role axioms. The third input parameter of a `&dlC’ atom specifies the ternary predicate name.
wine(X) :- &dlC[“wine.rdf”,a,b,l,d,”Wine”](X). l(locatedIn,X,Y) :- grows(X,Y).
In this program, before we query the concept `”Wine”’, we extend concept `locatedIn’ by predicate `grows’ with the aid of `l’.
Role querying can be accomplished using the external atom
(4) &dlR[KB,a,b,c,d,Q](X,Y),
where `Q’ is a role name and `X’ and `Y’ are terms. If the external atom has a non-ground output, i.e. both `X’ and `Y’ are variables, then (4) retrieves all known pairs of role `Q’. If both `X’ and `Y’ are individuals, then (4) holds iff `(X,Y)’ is an instance of role `Q’. If only one of `X’ and `Y’ is ground, then we retrieve all fillers for the respective individual in the role.
OWL Datatype Properties queries are subject to a different atom:
(5) &dlDR[KB,a,b,c,d,Q](X,Y)
is basically the same as the `&dlR’ atom, but `Q’ is a Datatype Property here.
The input mechanism of the `&dlC’ external atom and the role querying atoms is the same.
Example 4: nzwine(X) :- &dlR[“wine.rdf”,a,l,c,d,locatedIn](X,”NewZealandRegion”). l(locatedIn,X,Y) :- grows(X,Y).
The next two atoms provide support for conjunctive and union of conjunctive queries:
(6) &dlCQ[KB,a,b,c,d,CQ](X1,…,Xn) and (7) &dlUCQ[KB,a,b,c,d,UCQ](X1,…,Xn),
where `CQ’ and `UCQ’ is a conjunctive query and a union of conjunctive queries, respectively. Both output lists consists of an n-ary tuple `(X1,…,Xn)’, where each `Xi’ is a variable. Note that in general, HEX-programs allow ground terms as arguments in external atoms, but ground terms in the output of `&dlCQ’ and `&dlUCQ’ atoms are useless. Therefore, we omit the case where some `Xi’ may be ground.
A conjunctive query `CQ’ is a query of form
(8) Q1(Y1),Q2(Y2),…,Qm(Ym),
where for 1 <= i <= m each `Qi’ is a concept or role name and `Yi’ a single term or a pair of terms.
Similarly, a union of conjunctive queries `UCQ’ is a query of form
(9) CQ1 v CQ2 v … v CQn,
where for 1 <= i <= n each `CQi’ is a conjunctive query.
Example 5: Consider the following rule issuing a conjunctive query over the wine ontology. As result, we retrieve all dry wines from the `Burgundy’ region in predicate `bdw’.
#namespace(vin,”http://www.w3.org/TR/2003/PR-owl-guide-20031209/wine#”). bdw(X) :- &dlCQ[“wine.rdf”,a,b,c,d,”Burgundy(X),hasSugar(X,vin:Dry)”](X).
Example 6: UCQs can be expressed in the following manner.
#namespace(food,”http://www.w3.org/TR/2003/PR-owl-guide-20031209/food#”). wp(X) :- &dlUCQ[“wine.rdf”,a,b,c,d,”WhiteWine(X) v food:Pasta(X)”](X).
The above rule retrieves all white wines and pasta dishes.
Again, the input mechanism of the `&dlC’ external atom and `&dlCQ’/`&dlUCQ’ is the same. Instead of `a’, `b’, `c’, `d’, we can use similar input predicates as in the Examples 2-4.
The last external atom is `&dlConsistent’, which tests the given DL-KB for consistency under the specified extensions.
(10) &dlConsistent[KB,a,b,c,d]
If `KB’ is consistent after possibly augmenting the ABox according to the input list, the atom (10) evaluates to true, otherwise false.
Example 7: The program
p(“Wine”,”Uhudler”). n(“Wine”,”Uhudler”). :- not &dlConsistent[“wine.rdf”,p,n,c,d].
has no answer set, since we extend the ABox of `wine.rdf’ by `Wine(Uhudler)’ and `-Wine(Uhudler)’.
In order to process cq-programs with dlvhex, we can write DL-atoms as
(11) DL[S1 op1 p1,…,Sm opm pm; q](X1,…,Xn)
where `q(X1,…,Xn)’ is a dl- or cq-query and `opi’ may be `+=’ and `-=’ for concept/role and negated concept/role augmentation, resp. Since we do not encode the ontology in the program, we must add the option `–ontology=URL’ to the `dlvhex’ command line arguments, where `URL’ is a file or a URL to an OWL Ontology.
Example 8: A simple example of a cq-program is (L,P), where L is the wine ontology in `wine.rdf’ and P is the following rule
wine(X) :- DL[Wine](X).
(L,P) is equivalent to the program in Example 1, provided that above rule is in a file called `wine.dlp’ and we call dlvhex as follows:
$ dlvhex –ontology=wine.rdf wine.dlp
Similarly, we can encode Example 2 as cq-program
wine(X) :- DL[Wine += w; Wine](X). w(“Uhudler”).
and get the same extensions in `wine’ as in the HEX-rule, given that we call the dlvhex program in the same way as above.
TODO