Status and News
Klaus D. Witzel
I wanted to report on the progress already last month, but a new customer
draws all my attention. It is quite unusual that in my shop we start a new
project in December/January but, we can't complain ;)
-------------------
As to the status of Mindlog project: I've made more than a dozen of heavy
tests (and some tweaks to the object model), which still await upload.
When I wrote the tests, a peculiar pattern emerged out of the WordNet
data, which still puzzles me, since I couldn't find any explanation. An
example of the pattern (lemma of WordNet in ''s):
'wing' relates to 'body part'
in the same ways as
'bird' relates to 'organism'
Note that ('wing' isPartOf 'bird') and ('body part' isPartOf 'organism'),
and that ('wing' isKindOf 'body part') and ('bird' isKindOf 'organism').
This gives rise to a "magic square" in which the "in the same way as"
relation, from the above example, seemingly emerges "out of the blue". I
shall upload a diagram which pictures this.
I've written many queries of the above sort, taking other basic WordNet
relations (other than isKindOf, isPartOf) into account as well, and the
"in the same way as" hidden relation of this "magic square" also appears
for
- isOppositeOf
- isAttributeOf
- isDerivationalOf
- isCauseOf
- isMemberOf
- etc
with varying # of occurences; most occurences can be found with
isDerivationalOf, 3nd most has isPartOf, followed by isOppositeOf.
-------------------
There is no mention of such a "hidden relation" in the papers on WordNet,
and I'm going to write the WordNet people about it.
I can only speculate that this "in the same way as" is a consequence of
their methodology when they created WordNet, or that this sort of "hidden
relation" reflects their intuition when choosing lemma/words and their
basic relations.
What fascinates me is that WordNet seemingly shows a way to learn
from/"by" itself, that would be fantastic ;)
Cheers,
Klaus
"If at first, the idea is not absurd, then there is no hope for it".
Albert Einstein