Search code examples
nlpwordnet

What are "derivationally related forms" in WordNet?


WordNet lemmas can have derivationally related forms. For instance, the noun "butter" (meaning the spread you put on bread) is deemed to be derivationally-related to the verb "butter" (meaning the act of spreading butter on bread):

>>> from nltk.corpus import wordnet as wn
>>>
>>> wn.lemma('butter.n.01.butter')
Lemma('butter.n.01.butter')
>>> wn.lemma('butter.n.01.butter').synset().definition()
u'an edible emulsion of fat globules made by churning milk or cream; for cooking and table use'
>>> wn.lemma('butter.n.01.butter').derivationally_related_forms()
[Lemma('butter.v.01.butter'), Lemma('buttery.s.02.buttery'), Lemma('butyraceous.a.01.butyraceous')]
>>> wn.lemma('butter.n.01.butter').derivationally_related_forms()[0]
Lemma('butter.v.01.butter')
>>> wn.lemma('butter.n.01.butter').derivationally_related_forms()[0].synset().definition()
u'spread butter on'
>>>
>>> wn.lemma('flood.n.01.flood').synset().definition()
u'the rising of a body of water and its overflowing onto normally dry land'
>>> wn.lemma('flood.n.01.flood').derivationally_related_forms()
[Lemma('flood.v.04.flood'), Lemma('deluge.v.01.flood'), Lemma('flood.v.02.flood')]
>>> wn.lemma('flood.n.01.flood').derivationally_related_forms()[0]
Lemma('flood.v.04.flood')
>>> wn.lemma('flood.n.01.flood').derivationally_related_forms()[0].synset().definition()
u'become filled to overflowing'
>>>

However, it's not clear precisely what the term "derivationally related" means. For instance, I could argue that "television" and "telescope" are "derivationally related", since both words are derived from the Ancient Greek "têle", meaning "far". But WordNet disagrees:

>>> wn.lemma('telescope.n.01.telescope').derivationally_related_forms()
[Lemma('telescopic.s.01.telescopic'), Lemma('telescopic.s.02.telescopic')]

What precisely, then, is WordNet's definition of a "derivationally related form"? Is this documented anywhere?


Solution

  • From the WordNet glossary:

    derivationally related forms: Terms in different syntactic categories that have the same root form and are semantically related.

    The verb "to butter" has the same root form as the noun "butter", they are different syntactic categories (verb vs. noun), and they're clearly semantically related.

    On the other hand, "television" and "telescope" do not have the same root form (but e.g. "to televise" would have the same root as "television", and "telescopic" has the same root as "telescope"). Moreover, they are not semantically related, and they are both nouns.