Using UMBEL in semantic similarity measure

[Nasr Eddine]

Hi,

I want to use UMBEL to calculate similarity value between concepts. So, I need a taxonomy of concepts.

 I've used skos:broaderTransitive property to construct hierarchical taxonomic between reference concepts.

Unfortunately, I found that the constructed graph is not DAG (Direct Acyclic Graph).

What do you suggest to convert constructed graph into DAG ?

I found also that there are many relations that could be used to construct concepts taxonomy (like rdfs:subClassOf, skos:broader ...), what the better relation to consider ?

Kind regards.

[Frederick Giasson]

Hi!

> I want to use UMBEL to calculate similarity value between concepts.
> So, I need a taxonomy of concepts.

Good

> I've used skos:broaderTransitive property to construct hierarchical
> taxonomic between reference concepts.
> Unfortunately, I found that the constructed graph is not DAG (Direct
> Acyclic Graph).

Yeah, there are cycles in the graph.

> What do you suggest to convert constructed graph into DAG ?

You would have to detect the cycles and to remove the broaderTransitive
triple that creates that cycle.

> I found also that there are many relations that could be used to
> construct concepts taxonomy (like rdfs:subClassOf, skos:broader ...),
> what the better relation to consider ?

You should consider using skos:broader/narrower for that kind of work.

Hope it helps!

Thanks,

Fred

[Nasr Eddine]

Hi, 

Thank your for your replay. 

I'm little bit confused, here you recommended me to use skos:broader/narrower to construct the taxonomy while in your website you've recommended using skos:broaderTransitive. 

You meant by skos:broader is not transitive that we cannot have taxonomic hierarchy if we follow property path ?

Because I learnt from skos specification that we can contruct skos:broaderTransitive from skos:broader; I do not know how did you use these skos properties.

Thanks.

--
--
You received this message because you are subscribed to the "UMBEL" group.
To post:  umbel-o...@googlegroups.com
To unsubscribe:  umbel-ontolog...@googlegroups.com
Group home:  http://groups.google.com/group/umbel-ontology?hl=en
--- You received this message because you are subscribed to a topic in the Google Groups "UMBEL" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/umbel-ontology/Sb1CSBa9lR8/unsubscribe.
To unsubscribe from this group and all its topics, send an email to umbel-ontolog...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Frederick Giasson]

Hi,

I'm little bit confused, here you recommended me to use skos:broader/narrower to construct the taxonomy while in your website you've recommended using skos:broaderTransitive.

Sorry, I meant skos:broaderTransitive/narrowerTransitive

You meant by skos:broader is not transitive that we cannot have taxonomic hierarchy if we follow property path ?

No, that is right, you cannot create it with it.

Because I learnt from skos specification that we can contruct skos:broaderTransitive from skos:broader; I do not know how did you use these skos properties.

The skos:broaderTransitive vs. skos:broader depends on how we converted the OpenCyc concept into UMBEL reference concepts. The transitive structure is really defined using skos:broaderTransitive *and not* skos:broader. The skos:broader predicate comes from the rdf:type relationship of named individuals in OpenCyc, which doesn't define a transitive structure.


Hope it helps clarifying these points.

Thanks,

Fred

[hbutt.ms...@seecs.edu.pk]

Hi, I have recently joined this group and looking for the same question. Can you please share approach? Regards

[Frederick Giasson]

Hi,

rdfs:subClassOf is probably the best path to take. To remove the cycles,
you should first list each of them and then remove the subClassOf
relation that creates the cycle.

Hope it helps,

Thanks,

Fred