SuperTypes and Cyc mapping
Aleksander Pohl
First of all, I would like to express my appreciation for the idea of
SuperTypes introduced to UMBEL. I think this is one of the things which
was missing in Cyc, which could substantially smoothen its learning
curve and broaden its adoption. Great work!
Still, I have two observations with respect to UMBEL 1.0:
1. I downloaded the super_types.zip containing the partitioning of
concepts into super_types - there is at least one error in these files,
namely the file superTypes_Activities.csv has one owl:Nothing entry. I
did not look into other files, but I suspect there might more errors of
this type.
2. The names of the reference concepts are sometimes different from the
names in Cyc. The "-" sign is substituted with "_", but there are also
some concepts (I believe) not present both in OpenCyc nor in Research
Cyc (e.g. Freeware_game, Cherry, etc.). So first, I am curious to know,
if the substitution scheme (_ -> -) is correct, and second where does
these extra-concepts come from.
Kind regards,
Aleksander Pohl
Frederick Giasson
> First of all, I would like to express my appreciation for the idea of
> SuperTypes introduced to UMBEL. I think this is one of the things which
> was missing in Cyc, which could substantially smoothen its learning
> curve and broaden its adoption. Great work!
Thanks,
> Still, I have two observations with respect to UMBEL 1.0:
> 1. I downloaded the super_types.zip containing the partitioning of
> concepts into super_types - there is at least one error in these files,
> namely the file superTypes_Activities.csv has one owl:Nothing entry. I
> did not look into other files, but I suspect there might more errors of
> this type.
Well, this is not really an error. This is an outcome of the usage of
the OWLAPI to generate these lists. True that we should filter it out
for the next version, but this has no impact on anything within the list.
However, why is that? This is really simple. In OWL, owl:Nothing is a
subClassOf *all* possible classes. So, this is why it is there, and why
it has no impact.
> 2. The names of the reference concepts are sometimes different from the
> names in Cyc. The "-" sign is substituted with "_", but there are also
> some concepts (I believe) not present both in OpenCyc nor in Research
> Cyc (e.g. Freeware_game, Cherry, etc.). So first, I am curious to know,
> if the substitution scheme (_ -> -) is correct, and second where does
> these extra-concepts come from.
About Freeware_game, here is the OpenCyc concept [1]. The "human
readable URI" may have changed between opencyc versions; however, the
linkage between different OpenCyc versions is done using Cyc's unique,
internal ID (such as
"Mx8Ngh4rvyz3xZwpEbGdrcN5Y29ycB4rvVjmuJwpEbGdrcN5Y29ycA") and not the
human readable label.
So, the concepts should be there even if not under the same human
readable form.
About the "_" and "-", to my memory, everything was underscored in
OpenCyc. Or it may be a difference between versions. Would need some
examples for this one.
[1]
http://sw.opencyc.org/concept/Mx8Ngh4rvyz3xZwpEbGdrcN5Y29ycB4rvVjmuJwpEbGdrcN5Y29ycA
Hope this helps
Thanks!
Take care,
Fred
> Kind regards,
> Aleksander Pohl
>
Aleksander Pohl
>> Still, I have two observations with respect to UMBEL 1.0:
>> 1. I downloaded the super_types.zip containing the partitioning of
>> concepts into super_types - there is at least one error in these files,
>> namely the file superTypes_Activities.csv has one owl:Nothing entry. I
>> did not look into other files, but I suspect there might more errors of
>> this type.
>
> Well, this is not really an error. This is an outcome of the usage of the
> OWLAPI to generate these lists. True that we should filter it out for the
> next version, but this has no impact on anything within the list.
>
> However, why is that? This is really simple. In OWL, owl:Nothing is a
> subClassOf *all* possible classes. So, this is why it is there, and why it
> has no impact.
>
>> 2. The names of the reference concepts are sometimes different from the
>> names in Cyc. The "-" sign is substituted with "_", but there are also
>> some concepts (I believe) not present both in OpenCyc nor in Research
>> Cyc (e.g. Freeware_game, Cherry, etc.). So first, I am curious to know,
>> if the substitution scheme (_ -> -) is correct, and second where does
>> these extra-concepts come from.
>
> About Freeware_game, here is the OpenCyc concept [1]. The "human readable
> URI" may have changed between opencyc versions; however, the linkage between
> different OpenCyc versions is done using Cyc's unique, internal ID (such as
> "Mx8Ngh4rvyz3xZwpEbGdrcN5Y29ycB4rvVjmuJwpEbGdrcN5Y29ycA") and not the human
> readable label.
Sure, I wouldn't ask, if I had found the mapping between Umbel
concepts and Cyc concepts based on the external IDs. But I haven't
found them (at least they are not present in the
umbel_reference_concepts_v100.n3 file or my Virtuous did not report
them).
Kind regards,
Aleksander
Frederick Giasson
>> Well, this is not really an error. This is an outcome of the usage of the
>> OWLAPI to generate these lists. True that we should filter it out for the
>> next version, but this has no impact on anything within the list.
>>
>> However, why is that? This is really simple. In OWL, owl:Nothing is a
>> subClassOf *all* possible classes. So, this is why it is there, and why it
>> has no impact.
> Ok. Just wanted to inform you about the issue, even though it has no impact :)
Sure thanks!
>>> 2. The names of the reference concepts are sometimes different from the
>>> names in Cyc. The "-" sign is substituted with "_", but there are also
>>> some concepts (I believe) not present both in OpenCyc nor in Research
>>> Cyc (e.g. Freeware_game, Cherry, etc.). So first, I am curious to know,
>>> if the substitution scheme (_ -> -) is correct, and second where does
>>> these extra-concepts come from.
>> About Freeware_game, here is the OpenCyc concept [1]. The "human readable
>> URI" may have changed between opencyc versions; however, the linkage between
>> different OpenCyc versions is done using Cyc's unique, internal ID (such as
>> "Mx8Ngh4rvyz3xZwpEbGdrcN5Y29ycB4rvVjmuJwpEbGdrcN5Y29ycA") and not the human
>> readable label.
> Sure, I wouldn't ask, if I had found the mapping between Umbel
> concepts and Cyc concepts based on the external IDs. But I haven't
> found them (at least they are not present in the
> umbel_reference_concepts_v100.n3 file or my Virtuous did not report
> them).
Hummm right, it is not available as we speak. I will check to add it to
the SVN as well (it is another CSV file used in our umbel generation
scripts; this is a series of scripts used to generate new version of
UMBEL based on new versions of OpenCyc).
Thanks!
Fred