RdfGraph schema retrieval queries for the relation types are not linked by the correct comment variable

System Info

langchain = 0.0.251 Python = 3.10.11

Who can help?

No response

Information

[ ] The official example notebooks/scripts
[X] My own modified scripts

Related Components

[ ] LLMs/Chat Models
[ ] Embedding Models
[ ] Prompts / Prompt Templates / Prompt Selectors
[ ] Output Parsers
[ ] Document Loaders
[ ] Vector Stores / Retrievers
[ ] Memory
[ ] Agents / Agent Executors
[ ] Tools / Toolkits
[X] Chains
[ ] Callbacks/Tracing
[ ] Async

Reproduction

Create an OWL ontology called dbpedia_sample.ttl with the following:


@prefix rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#> .
@prefix dcterms: <http://purl.org/dc/terms/> .
@prefix wikidata: <http://www.wikidata.org/entity/> .
@prefix owl: <http://www.w3.org/2002/07/owl#> .
@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> .
@prefix prov: <http://www.w3.org/ns/prov#> .
@prefix : <http://dbpedia.org/ontology/> .

:Actor a owl:Class ; rdfs:comment "An actor or actress is a person who acts in a dramatic production and who works in film, television, theatre, or radio in that capacity."@en ; rdfs:label "actor"@en ; rdfs:subClassOf :Artist ; owl:equivalentClass wikidata:Q33999 ; prov:wasDerivedFrom http://mappings.dbpedia.org/index.php/OntologyClass:Actor .

:AdministrativeRegion a owl:Class ; rdfs:comment "A PopulatedPlace under the jurisdiction of an administrative body. This body may administer either a whole region or one or more adjacent Settlements (town administration)"@en ; rdfs:label "administrative region"@en ; rdfs:subClassOf :Region ; owl:equivalentClass http://schema.org/AdministrativeArea, wikidata:Q3455524 ; prov:wasDerivedFrom http://mappings.dbpedia.org/index.php/OntologyClass:AdministrativeRegion .

:birthPlace a rdf:Property, owl:ObjectProperty ; rdfs:comment "where the person was born"@en ; rdfs:domain :Animal ; rdfs:label "birth place"@en ; rdfs:range :Place ; rdfs:subPropertyOf dul:hasLocation ; owl:equivalentProperty http://schema.org/birthPlace, wikidata:P19 ; prov:wasDerivedFrom http://mappings.dbpedia.org/index.php/OntologyProperty:birthPlace .


2. Run
``` python
from langchain.graphs import RdfGraph

graph = RdfGraph(
    source_file="dbpedia_sample.ttl",
    serialization="ttl",
    standard="owl"
)

print(graph.get_schema)

Output

In the following, each IRI is followed by the local name and optionally its description in parentheses. 
The OWL graph supports the following node types:
<http://dbpedia.org/ontology/Actor> (Actor, An actor or actress is a person who acts in a dramatic production and who works in film, television, theatre, or radio in that capacity.),
<http://dbpedia.org/ontology/AdministrativeRegion> (AdministrativeRegion, A PopulatedPlace under the jurisdiction of an administrative body. This body may administer either a whole region or one or more adjacent Settlements (town administration))
The OWL graph supports the following object properties, i.e., relationships between objects:
<http://dbpedia.org/ontology/birthPlace> (birthPlace, An actor or actress is a person who acts in a dramatic production and who works in film, television, theatre, or radio in that capacity.),
<http://dbpedia.org/ontology/birthPlace> (birthPlace, A PopulatedPlace under the jurisdiction of an administrative body. This body may administer either a whole region or one or more adjacent Settlements (town administration)), <http://dbpedia.org/ontology/birthPlace> (birthPlace, where the person was born)
The OWL graph supports the following data properties, i.e., relationships between objects and literals:

Expected behavior

The issue is that in the SPARQL queries getting the properties the rdfs:comment triple pattern always refers to the variable ?cls which obviously comes from copy/paste code.

For example, getting the RDFS properties via

rel_query_rdf = prefixes["rdfs"] + (
    """SELECT DISTINCT ?rel ?com\n"""
    """WHERE { \n"""
    """    ?subj ?rel ?obj . \n"""
    """    OPTIONAL { ?cls rdfs:comment ?com } \n"""
    """}"""
)

you can see that the OPTIONAL clause refers to ?cls, but it should be ?rel.

The same holds for all other queries regarding properties.

The current status leads to a cartesian product of properties and all rdfs:comment vlaues in the dataset, which can be horribly large and of course leads to misleading and huge prompts (see the output of my sample in the "reproduction" part)

langchain-ai / langchain