MaRQ: Mapping RDF Queries

This project is an approach to generate federated queries from RDF mappings.

Installation

We assume that Python3 and pip is installed

Download IDE for python

this project was created with Pycharm

Installation in a virtualenv is strongly recommended

install venv

python3 -m venv env

Install dependencies with pip:

pip install -r requirements.txt

Mappings

Mappings need to be Yarrrml mappings.

MaRQ use types to make queries, so a saturated mapping can create queries that would have been missed with a standard mapping.

'predicateobjects' should be of the form:

- [{one and only one predicate}, {one and only one object}]

Types or language aren't supported and might cause issues.

Run script

Run the following command to execute MaRQ on each possible pair from n mappings:

python script.py path/to/mapping/1 ... path/to/mapping/n

You can also specify a directory containing only mappings to execute MaRQ on every pair of mappings contained in the directory:

python script.py path/to/directory

It's also possible to specify the threshold at which a query is made from two template. Only a pair of template that have a Jaccard index >= of this value will pass.

python script.py float_value path/to/directory

Example

This example is made using the mappings [email protected] and [email protected].

Here are examples of query formed with those mappings. Each of those first features comments that indicates meta information on the queries:

• the type of join
• the mappings
• Which template of which mapping was used, and in which position
• The jaccard index number, which tells us how much the two template were similar. This index is calculated using the types link to a template, we consider both sets of types (one for each template) and then calculate the Jaccard index

### subject-subject
#M1: small-air-bnb-listings
#M2: small-annuaire-des-professionnels-de-sante
#M1_Subject :   Host
#M2_Subject :   Docteur
#Jaccard index: 0.3333333333333333
Select Count(?S2) Where {
        ?s2 rdf:type <http://schema.org/Person>.        #M1 M2
        ?s2 rdf:type <http://purl.org/goodrelations/v1#BusinessEntity>. #M1
        ?s2 rdf:type <http://linkedgeodata.org/ontology/Doctor>.        #M2
        ?s2 <https://schema.org/location> ?o1.  #M2
        ?s2 <http://dbpedia.org/ontology/speciality> ?o2.       #M2
}

This is a "subject-subject" query, which means the two templates that were used to join the mappings were both subjects in their respective mappings.

### object-object
#M1: small-air-bnb-listings
#M2: small-annuaire-des-professionnels-de-sante
#M1_Object :    Neighbourhood
#M2_Object :    Address
#Jaccard index: 0.3333333333333333
Select Count(?O2) Where {
       ?s1 <https://schema.org/containedInPlace> ?o2.  #M1
       ?s2 <https://schema.org/location> ?o2.  #M2
}

This is a "object-object" query, which means the two templates that were used to join the mappings were both objects in their respective mappings.

### subject-object
#M1: small-air-bnb-listings
#M2: small-annuaire-des-professionnels-de-sante
#M1_Object :    Host
#M2_Subject :   Docteur
#Jaccard index: 0.3333333333333333
Select Count(?T1) Where {
       ?t1 rdf:type <http://schema.org/Person>.        #M2
       ?t1 rdf:type <http://linkedgeodata.org/ontology/Doctor>.        #M2
       ?t1 <https://schema.org/location> ?f1.  #M2
       ?t1 <http://dbpedia.org/ontology/speciality> ?f2.       #M2
       ?f3 <http://dbpedia.org/ontology/owner> ?t1.    #M1
}

This is a "subject-object" query, which means the one of the templates that were used to join the mappings was a subject in its mapping and the other template was a object in hin the other a object

### subject-object
#M1: small-air-bnb-listings
#M2: small-annuaire-des-professionnels-de-sante
#M1_Subject :   Neighbourhood
#M2_Object :    Address
#Jaccard index: 0.3333333333333333
Select Count(?T1) Where {
       ?t1 rdf:type <http://schema.org/Place>. #M1
       ?t1 rdf:type <http://dbpedia.org/ontology/PopulatedPlace>.      #M1
       ?f1 <https://schema.org/location> ?t1.  #M2
}

This is another "subject-object" query, this time the subject is in mapping1 and the object is in mapping2