-
Notifications
You must be signed in to change notification settings - Fork 0
Wordnet
Wordnet 3.1 comes with textual db files in custom text format, see WordNet manual for reference. They can hold several semantic relations, but the most important ones for us are the commonly called IS-A and PART-OF, which we now describe in detail. Some good explanation can also be found here.
The IS-A relation (for example collie IS-A shepherd_dog), in linguistic terms can be also rephrased as collie hasHypernym shepherd_dog. Specifically, in WordNet we can find written a relation called hypernym (which uses @ as shorthand), that can be actually read like hasHypernym. So in the db you would find the edge
collie hypernym shepherd_dog
stated in a format like:
02106030 05 n 01 collie 0 001 @ 02104523 n 0000 | a silky-coated sheepdog with a long ruff and long narrow head developed in Scotland
where 02104523 is the ID of shepherd_dog
The inverse of hypernym would be hyponym (with shorthand ~)
Note original Wordnet db files have BOTH hypernym and hyponym edges.
There is also the special case for when an instance IS-A somenthing, like Italy IS-A European_Country. In linguistic terms can be also rephrased as Italy hasInstanceHypernym European_country. Specifically, in WordNet we can find written a relation called instance hypernym (which uses @i as shorthand), that can be actually read like hasInstanceHypernym. So in the db you would find the edge
Italy instanceHypernym European_country
stated in a format like:
08819530 15 n 03 Italy 0 Italian_Republic 0 Italia 0 078 @i 08714745 n 0000 #p 09298379 n 0000 #m 08190414 n 0000 #m 08191297 n 0000 + 02968612 a 0101 -r 01154838 a 0000 -r 01276493 n 0000 -r 01276664 n 0000 -r 01285678 n 0000 -r ...<REDACTED>.. | a republic in southern Europe on the Italian Peninsula; was the core of the Roman Republic and the Roman Empire between the 4th century BC and the 5th century AD
where 08714745 is the ID of European_country
The inverse of instance hypernym would be instance hyponym (with shorthand ~i)
Note original Wordnet db files have BOTH instance hypernym and instance hyponym edges.
The PART-OF relation (for example arm PART-OF armchair) in linguistic terms can be also be rephrased as arm hasPartHolonym armchair. Specifically, in WordNet we can find written three kind of PART-OF relations, part holonym, member holonym and substance holonym. We will focus on the only transitive one, part holonym. It uses #p as shorthand, and can be actually read like hasPartHolonym. So in the db you would find the edge
arm part holonym shepherd_dog
stated in a format like:
02737660 06 n 01 arm 2 003 @ 02741475 n 0000 #p 02738535 n 0000 ~ 04607982 n 0000 | the part of an armchair or sofa that supports the elbow and forearm of a seated person
where 02741475 is the ID of armchair.
The inverse of part holonym would be part meronym (with shorthand %p)
Note original Wordnet db files have BOTH part holonym and part meronym edges.
Uby converter 0.7.0 on Wordnet 3.0 produces these synset relations.
Notice that although antonym relations are present in Wordnet, they don't get imported!
usage
meronymSubstance
synonymNear
verbGroup
hyponymInstance
meronymMember
holonymPart
hypernymInstance
isUsageOf
topic
hyponym
holonymMember
region
isRegionOf
hypernym
holonymSubstance
entails
seeAlso
isTopicOf
attribute
causedBy
meronymPart