Allow prefixes in PropertyMapping

[gaurav]

A lot of the work is already done -- I've pulled it into my repo as https://github.com/gaurav/extraction-framework/tree/feature/prefix_in_propertymapping -- but it might need checking.

[jimkont]

Hey @ninniuz, @gaurav tries to merge your pull request, can you take a first review?

[ninniuz]

I think this is basically what I put into the MR one year ago...

[gaurav]

@ninniuz: Yup, I haven't changed your code at all! I just pulled your code onto the end of the current dbpedia:master, so it can be automatically merged. It works locally, except for generating a duplicate triple: I added a "prefix" to http://mappings.dbpedia.org/index.php/Mapping_commons:NARA-image-full and got two triples:

• The correct one: (http://commons.dbpedia.org/resource/File:John_James_Audubon_-_NARA_-_528730.jpg) (http://dbpedia.org/ontology/hasNationalArchivesIdentifier) "http://research.archives.gov/description/528730"^^(http://www.w3.org/2001/XMLSchema#string) .
• An odd duplicate: (http://commons.dbpedia.org/resource/File:John_James_Audubon_-_NARA_-_528730.jpg) (http://dbpedia.org/ontology/hasNationalArchivesIdentifier) "http://research.archives.gov/description/"^^(http://www.w3.org/2001/XMLSchema#string) .

We need to switch those triples to URIs and figure out why they're duplicating, but hopefully that won't take too long. Would you like to work on this or should I add it to my Google Summer of Code work?

[ninniuz]

@gaurav please go ahead :)

[gaurav]

Oops: the duplicate and the ^^xsd:string were both a result of hasNationalArchivesIdentifier being a datatype property. Once I converted it into an object property, it turned into a URI and the duplicate went away. I've updated http://mappings.dbpedia.org/index.php/Template:PropertyMapping to display the prefix/suffix/transform and to include some basic documentation. I'll try to update some templates to use prefixes and suffixes: I've tried to set it up so that mappings containing these are categorized into http://mappings.dbpedia.org/index.php/Category:Mappings_using_prefix and http://mappings.dbpedia.org/index.php/Category:Mappings_using_suffix, but that doesn't seem to be working.

[ninniuz]

I can see the categorization working. Got it fixed?

[gaurav]

I didn't do anything ... it just fixed itself :-P

[gaurav]

There are four cases to test:

• Does the code work when the value is:
  • A string/number, or
  • A URI
• Does the code work when the RDF property being produced is:
  • A DatatypeProperty
  • An ObjectProperty

For each case, we need to test both prefix and suffix.

[gaurav]

Properties tested:

• prefix
  • NARA-image-full: hasNationalArchivesIdentifier (value: string, produced: ObjectProperty)
  • doi: cites (value: string, produced: DatatypeProperty)
• suffix
  • ''None''
• prefix+suffix
  • Chemical structure verified (value: string, producted: ObjectProperty)
  • Chemical structure verified (value: string, producted: DatatypeProperty)

[gaurav]

When would the value ever be a DatatypeProperty? I tried to do something with http://mappings.dbpedia.org/index.php/Mapping_commons:Doi, but I think the mapping to a DatatypeProperty (doi:10.123abc/d) is overridding the mapping to an OntologyProperty (http://dx.doi.org/wiki/10.123abc/d). Apart from units, I can't think of a case where something would need to be prefixed or suffixed to a produce a non-URI (although I'm sure I'll come up with something thirty seconds after posting this update :-P).

[jimkont]

I cannot think of one either atm :)
The thing is that we should test if it behaves correctly, if yes then
everything is good ;)
if not, we should either fix it or make the mapping invalid

For testing this you can try it out with any string datatype
e.g. dbo:longName, dbo:title or foaf:name, it doesn't have to make sense
for now, just see it it works
you can try mapping the same template property to

1. foaf:name with prefix
2. dbo:longName with suffix
3. dbo:title with both
   Find 3 object properties as well and define them in a similar way

On Wed, Jul 23, 2014 at 7:07 AM, Gaurav Vaidya notifications@github.com
wrote:

When would the value ever be a DatatypeProperty? I tried to do something
with http://mappings.dbpedia.org/index.php/Mapping_commons:Doi, but I
think the mapping to a DatatypeProperty (doi:10.123abc/d) is overridding
the mapping to an OntologyProperty (http://dx.doi.org/wiki/10.123abc/d).
Apart from units, I can't think of a case where something would need to be
prefixed or suffixed to a produce a non-URI (although I'm sure I'll come up
with something thirty seconds after posting this update :-P).

—
Reply to this email directly or view it on GitHub
#29 (comment)
.

Kontokostas Dimitris

[gaurav]

Incidentally, http://mappings.dbpedia.org/index.php/Mapping_commons:Doi didn't work locally, so I think there is a problem with mapping the same value in two different ways, or maybe just how I did the mapping there.

[jimkont]

you mean it works with a single mapping in all different ways but not if
you define 2 or more?
maybe "1" is the problem here, can you exclude it and try with id only?

On Sun, Jul 27, 2014 at 6:42 PM, Gaurav Vaidya notifications@github.com
wrote:

Incidentally, http://mappings.dbpedia.org/index.php/Mapping_commons:Doi
didn't work locally, so I think there is a problem with mapping the same
value in two different ways, or maybe just how I did the mapping there.

—
Reply to this email directly or view it on GitHub
#29 (comment)
.

Kontokostas Dimitris

[gaurav]

OKAY

• http://mappings.dbpedia.org/index.php?title=Mapping_commons:Doi&oldid=37561 tests prefixes and suffixes for ObjectProperty and DatatypeProperty -> https://gist.github.com/gaurav/06c25381e81de884e58f#file-doi-ttl
• http://mappings.dbpedia.org/index.php?title=Mapping_commons:Chemical_structure_verified&oldid=37560 tests prefixes and suffixes together -> https://gist.github.com/gaurav/06c25381e81de884e58f#file-chemstructure-ttl

I think that means we're good to go? I'll look at it again tomorrow and check.

[jimkont]

so it was it a bug before or a wrong mapping?

On Mon, Jul 28, 2014 at 12:51 PM, Gaurav Vaidya notifications@github.com
wrote:

OKAY

http://mappings.dbpedia.org/index.php?title=Mapping_commons:Doi&oldid=37561
tests prefixes and suffixes for ObjectProperty and DatatypeProperty ->
https://gist.github.com/gaurav/06c25381e81de884e58f#file-doi-ttl

http://mappings.dbpedia.org/index.php?title=Mapping_commons:Chemical_structure_verified&oldid=37560
tests prefixes and suffixes together ->
https://gist.github.com/gaurav/06c25381e81de884e58f#file-chemstructure-ttl

I think that means we're good to go? I'll look at it again tomorrow and
check.

—
Reply to this email directly or view it on GitHub
#29 (comment)
.

Kontokostas Dimitris

[gaurav]

Ah, right. No, we don't have a working example for just prefix and just suffix on ObjectProperty. I'll try again.

[gaurav]

Ah, I knew I'd done prefixed ObjectProperties somewhere:

• http://mappings.dbpedia.org/index.php?title=Mapping_commons:NARA-image-full&oldid=36827 tests prefixes for ObjectProperty -> https://gist.github.com/gaurav/06c25381e81de884e58f#file-nara-ttl

I've added a spurious suffix to an ObjectProperty at http://mappings.dbpedia.org/index.php/Mapping_commons:Flickr, but it's not working. I'll need to debug this further.

[gaurav]

@jimkont: Do you think it makes sense to add prefix/suffix code for ObjectProperties fas done in @2d6e3ff? It doesn't make much sense to have a prefix for URLs, but suffixes might make sense. Note that converting a text value into an ObjectProperty works fine (e.g. hasNationalArchivesIdentifier in http://mappings.dbpedia.org/index.php?title=Mapping_commons:NARA-image-full&oldid=36827): the problem is when the code can recognize a URL before it gets to us. @2d6e3ff fixes that, but I'm not sure if it's necessary. What do you think?

[jimkont]

Let's keep it I think. If it doesn't make sense, people won't use it.
Otherwise we should make the mapping invalid