Adding something like a SenseDefinition element to Sense?

[anasfkhan81]

Might seem redundant (given synset definition) but would be useful in the case where the definition is taken from a dictionary or some other pre-existing lexical resource (as is our case)

[fcbond]

I agree. In fact, I thought it was already there!

…

On Wed, 6 Jul 2022 at 12:34, Anas Fahad Khan ***@***.***> wrote: Might seem redundant (given synset definition) but would be useful in the case where the definition is taken from a dictionary or some lexical resource (as is our case) — Reply to this email directly, view it on GitHub <#65>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAIPZRRQ2YSTZLVLNBVL63TVSVOMFANCNFSM52ZIN3GQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

-- Francis Bond <https://fcbond.github.io/>

[jmccrae]

I am not so sure, as it seems to go against the idea of WordNet of using synsets as the main conceptual modelling. I also feel that it would create a lot of confusion.

Another solution for @anasfkhan81 is to move all definitions to the synset and use the Dublin Core properties to indicate how they were extracted from the dictionary.

Are there any existing wordnets that model definitions at the sense level?

[fcbond]

The Polish wordnet has definitions at the sense level.

…

On Wed, 6 Jul 2022 at 13:19, John McCrae ***@***.***> wrote: I am not so sure, as it seems to go against the idea of WordNet of using synsets as the main conceptual modelling. I also feel that it would create a lot of confusion. Another solution for @anasfkhan81 <https://github.com/anasfkhan81> is to move all definitions to the synset and use the Dublin Core properties to indicate how they were extracted from the dictionary. Are there any existing wordnets that model definitions at the sense level? — Reply to this email directly, view it on GitHub <#65 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAIPZRQXHJS2WA7QFSWJPL3VSVTU3ANCNFSM52ZIN3GQ> . You are receiving this because you commented.Message ID: ***@***.***>

-- Francis Bond <https://fcbond.github.io/>

[anasfkhan81]

From Globalex I know dictionary definitions are also a big part of the Turkish KeNet (where they use them to carry out diachronic research on sense changes) and in general, in the case a WN has been bootstrapped or derived from another lexical resource I definitely think the original definitions should be accessible somehow (whether from Synset or Sense) especially in the case of (semi)-automatic deriviation. From this pov i think they would count as an auxiliary part of the sense layer of a WN rather than part of the main conceptual modelling which is as always determined by synsets, their definitions and inter-relations.

[goodmami]

I'd agree with @jmccrae that definitions belong on synsets. A definition explains what something means and the synset is the locus of that conception. In contrast, we allow examples on senses because with an example you generally want to see the usage of a particular word with a given meaning (I'd even argue that examples should only appear on senses).

I appreciate the argument that the source material for a wordnet may be a dictionary where each word (and thus each synonym of some concept) has a definition, but I don't think a wordnet is necessarily a superset of a dictionary; i.e., that it should fully encode all information that a dictionary does. If it is important to capture the definitions of all the words, a synset definition with dc:source, as @jmccrae suggested, seems adequate.