Semantic mapping for Knowledge Graph visualization

The main goal of this project is to automate the process of creating semanitc maps to visualize knowledge graphs.

Sematch installation

In order to compute the semantic similarity for each node in the KG, we will use the sematch python library.

Sematch installation

The current released version of sematch library does not support Python 3.9. To use a python 3+ compatible version, we install the code from branch ´py3compat´.

git clone https://github.com/gsi-upm/sematch.git
cd sematch
git checkout py3compat
git pull origin py3compat
python setup.py install

Dataset generation

In order to test these modules, we need to provide a Knowledge Graph encoded as a list of N-triples (.nt files). To generate these datasets, we used and recommend to use the DBPedia endpoint to run SPARQL queries. This endpoint offers the capability to generate N-triples files.

In the following subsections, we present the SPARQL queiries used to generate the datasets used to validate the functionality of the summarization process to visualize Knwoledge Graphs.

SCI-FI-MOVIES.NT

PREFIX yago: <http://dbpedia.org/class/yago/>
PREFIX dbp: <https://dbpedia.org/property/>
PREFIX xsd: <http://www.w3.org/2001/XMLSchema#>
CONSTRUCT { 
    ?movie rdf:type  yago:Movie106613686 .
}
WHERE {
    ?movie rdf:type yago:Movie106613686 ;
               dbo:wikiPageWikiLink  dbc:American_science_fiction_action_films ;
               dbo:gross ?gross .
    FILTER(?gross > 8E8)
}

FANTASY-NOVELS.NT

PREFIX yago: <http://dbpedia.org/class/yago/>
PREFIX dbp: <https://dbpedia.org/property/>
PREFIX xsd: <http://www.w3.org/2001/XMLSchema#>
CONSTRUCT { 
    ?book rdf:type  yago:FWikicatFantasyNovels .
}
WHERE {
    ?book rdf:type yago:WikicatFantasyNovels .
    FILTER ( dbp:published > "2000-1-1"^^xsd:date  )
}

CITIES.NT

CONSTRUCT { 
?city rdf:type  yago:City108524735 .
}
WHERE {
?city rdf:type yago:City108524735 ;
            dbo:populationTotal ?populationTotal .
    FILTER(?populationTotal > 5E6 )
}

DISEASES.NT

PREFIX yago: <http://dbpedia.org/class/yago/>
CONSTRUCT { 
    ?disease rdf:type  yago:Disease114070360 .
}
WHERE {
   ?disease  rdf:type yago:Disease114070360 ;
            dbo:wikiPageWikiLink dbr:Infectious_disease .
}

DRUGS.NT

PREFIX yago: <http://dbpedia.org/class/yago/>
CONSTRUCT { 
    ?med rdf:type yago:Compound114818238 .
}
WHERE {
   ?disease  rdf:type yago:Disease114070360 ;
            dbo:medication ?med .
   ?med rdf:type yago:Compound114818238 ;
                       rdf:type dbo:Drug .
}

ACTORS.NT

PREFIX dbo: <http://dbpedia.org/ontology/>
PREFIX dbp: <http://dbpedia.org/property/>
CONSTRUCT {
    ?actor rdf:type yago:Actor109765278 .
}
WHERE {
    ?movie rdf:type yago:Movie106613686 ;
        dbo:wikiPageWikiLink  dbc:American_science_fiction_action_films ;
        dbo:starring ?actor .
?actor rdf:type yago:Actor109765278 .
} LIMIT 400s

Unit Test execution

In order to test the semantic mapping process, the test suite under test folder contains a set of Unit Tests that read some pre-computed sematic similarity matrices and generates the semantic map for them.

To run this suite simply go to test folder and run the following command:

python -m pytest

It will run all UTs and generate the corresponding semantic maps.

Example of UTs output

========== test session starts =====================================================
platform darwin -- Python 3.9.7, pytest-6.2.4, py-1.10.0, pluggy-0.13.1
rootdir: /Users/pcamarillor/Code/owl_python/semantic_mapping/test
plugins: anyio-2.2.0
collected 4 items                                                                                                            

semantic_similarity_test.py ....                                                                                       [100%]

================================================ 4 passed in 91.32s (0:01:31) ================================================