Issue 149 in uby: import of sense alignmnents using temporary tables sometimes fails

[u...@googlecode.com]

Status: Accepted
Owner: eckle.kohler
Labels: Type-Defect Priority-Medium Milestone-0.8.0

New issue 149 by eckle.kohler: import of sense alignmnents using temporary
tables sometimes fails
https://code.google.com/p/uby/issues/detail?id=149

for example when using
GermaNetWiktionaryDeAlignment

and adapting it to querying a database containing OntoWiktionary instead of
Wiktionary (different Uby sense IDs, but same original sense IDs).

Problems:
1) all pairs in SenseAxis are wrong
2) for some strange reason, the import script also pairs senses which are
both from OntoWiktionary

example:

ad 1) <SenseAxis id="GN9_OntoWktDE_14"
senseOne="GN_Sense_22345" senseTwo="OntoWktDE_sense_5671"
senseAxisType="monolingualSenseAlignment"/>
-> Misskredit, Blockflöte

<SenseAxis id="GN9_OntoWktDE_16"
senseOne="GN_Sense_18545" senseTwo="OntoWktDE_sense_53381"
senseAxisType="monolingualSenseAlignment"/>
-> Therapieform, Blockhaus

ad 2) <SenseAxis id="GN9_OntoWktDE_15"
senseOne="OntoWktDE_sense_12115" senseTwo="OntoWktDE_sense_5671"
senseAxisType="monolingualSenseAlignment"/>




--
You received this message because this project is configured to send all
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings

[u...@googlecode.com]

Updates:
Labels: Module-integration.alignment

Comment #1 on issue 149 by chmeyer.de: import of sense alignmnents using

temporary tables sometimes fails
https://code.google.com/p/uby/issues/detail?id=149

One crucial problem is that OntoWiktionary != Wiktionary. So far, we keep
the old 2011 dump version of Wiktionary around, mainly because we haven't
replaced the original word sense alignment I've created in 2011 with a
newer one based on DWSA. OntoWiktionary, however, makes use of a 2013 dump
and uses a different JWKTL version. Thus, the original sense IDs are NOT
compatible. This should explain why all SenseAxis pairs are wrong (Sorry, I
could have raised this earlier, but I thought that using the new alignment
framework we had newly created, OntoWiktionary-specific alignments).

This of course does not explain why in some cases two OntoWiktionary senses
are aligned. I cannot say much about that, but probably there is a lexicon
check missing? It is possible that an original sense ID of OntoWiktionary
and matches an original ID from a different resource. It is therefore
crucial to check the lexicon (respetively, the external system identifer).
If that's not the issue, than there's of course the chance for a major bug
in the software - I did not check any source code before filling up this
textarea...

[u...@googlecode.com]

Comment #2 on issue 149 by eckle.kohler: import of sense alignmnents using

temporary tables sometimes fails
https://code.google.com/p/uby/issues/detail?id=149

>> OntoWiktionary, however, makes use of a 2013 dump and uses a different
>> JWKTL version.
>> Thus, the original sense IDs are NOT compatible.

I am aware of that. Yet, the alignment of _the original sense IDs_ appears
to be still (mostly I guess) valid (I did not find a wrong alignment yet
when hand picking arbitrary pairs and looking them up via their
MonolingualExternalRefs) iff it is imported via the Uby API (which checks
for original sense ID AND external system).

However the import via temporary tables fails as described. This might
indeed be caused by not checking for expernalSystem - but the database that
I used for looking up original sense IDs contained only GermaNet, WordNet
and OntoWiktionary - no other lexicon.