Why adding transitivity breaks language filter in SPARQL Query?

sparql_query_inference

@prefix tpedia: <http://tatipedia.org/> .
@prefix owl: <http://www.w3.org/2002/07/owl#> .
@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> .

tpedia:new_york a tpedia:Place ;
                rdfs:label "Nova Iorque"@pt ;
                rdfs:label "New York"@en .

tpedia:london a tpedia:Place ;
                rdfs:label "Londres"@pt ;
                rdfs:label "London"@en .

tpedia:name rdf:type owl:DatatypeProperty ;
            rdfs:subPropertyOf rdfs:label .

tpedia:munich a tpedia:Place ;
                tpedia:name "Munique"@pt ;
                tpedia:name "Munich"@en .

I'd like to build a query to retrieve a list of instances' URIs and rdfs:labels provided:

• class (tpedia:Place)
• language filter (e.g. "pt" ou "en")

For this task I have loaded the TTL above into OpenLink Virtuoso OpenSource.

The query bellow:

SELECT DISTINCT ?subject ?label
WHERE {
    ?subject a <http://tatipedia.org/Place>;
             rdfs:label ?label .
    FILTER(langMatches(lang(?label), "pt")) .
}

Returns for this dataset:

+-----------------+------------------+
|     subject     |      label       |
+-----------------+------------------+
| tpedia:london   | "Londres"@pt     |
| tpedia:new_york | "Nova Iorque"@pt |
+-----------------+------------------+

Ignoring:

+---------------+--------------+
|    subject    |    label     |
+---------------+--------------+
| tpedia:munich | "Munique"@pt |
+---------------+--------------+

While my intention is:

+-----------------+------------------+
|     subject     |      label       |
+-----------------+------------------+
| tpedia:london   | "Londres"@pt     |
| tpedia:munich   | "Munique"@pt     |
| tpedia:new_york | "Nova Iorque"@pt |
+-----------------+------------------+

I tried to fix this problem using transitivity:

SELECT DISTINCT ?subject ?label
WHERE {
    ?subject a <http://tatipedia.org/Place>;
             rdfs:label ?label option (transitive, t_distinct, t_step('step_no') as ?n, t_min(0)) .
    FILTER (langMatches(lang(?label), "pt"))
}

But it returned:

+--------------------------------+-------------------------------+
|            subject             |             label             |
+--------------------------------+-------------------------------+
| http://tatipedia.org/new_york  | http://tatipedia.org/new_york |
| http://tatipedia.org/new_york  | "New York"@en                 |
| http://tatipedia.org/new_york  | "Nova Iorque"@pt              |
| http://tatipedia.org/london    | http://tatipedia.org/london   |
| http://tatipedia.org/london    | "London"@en                   |
| http://tatipedia.org/london    | "Londres"@pt                  |
| http://tatipedia.org/munich    | http://tatipedia.org/munich   |
+--------------------------------+-------------------------------+

Not even applying filters.

Any suggestions?

The issue was solved using inference.

First, we need to add the inference rule using the Virtuoso ISQL:

rdfs_rule_set ('http://tatipedia.org/property_ruleset', 'http://tatipedia.org');

After this, the query bellow will work as expected:

DEFINE input:inference <http://tpedia.org/property_ruleset>
SELECT DISTINCT ?subject ?label
WHERE {
    ?subject a <http://tatipedia.org/Place>;
             rdfs:label ?label  .
    FILTER (langMatches(lang(?label), "pt"))
}