Universal Networking Language seems to be most advanced system from the list https://meta.wikimedia.org/wiki/Abstract_Wikipedia/Related_and_previous_work/Natural_language_generation
Its main advantage is that it is not lemma-oriented but sense-oriented.
US patents
6,704,700 – active, 2020-11-17 Anticipated expiration
7,107,206 – Status Expired - Fee Related, 2022-07-22 Adjusted expiration
can we register to http://www.unlweb.net/unlweb/ and for example get some resources
see also github page: https://github.com/dikonov/Universal-Dictionary-of-Concepts
Main dictionary is sense list , one sense can have several lemmas synonyms in many languages.
Definitions is UNL-ized
I think, it will be well if Abstract would inspired by UNL. If it will possible if sense definitions will not only in Abstract graph, but full understood by computer? , for example:
https://en.wiktionary.org/wiki/star#Noun has 11 definitions,
1 and 2 are near, 4 and 5 are near, 1 and 4 are far, 4 seems to be sub-definition of 5
it is possible that computer find this properties of definitions?
Also each definition might have some context example and even distinguish rules (for WSD):
def 4: person, activities on film
def 5, person, general activities
def 9: often in phrase: his,her,out,their star
etc
I think it might several spaces:
- uniwersal dictionary of senses with definitions in abstract graph
- dictionaries for English and other languages
- main space – not common words but encyclopedic entities
Advantages of UNL- like approach
- sense-oriented
- can generate text in natural languages
- resource of understanding definitions
- data for learning word sense disambiguation problem
Best regards,
Andrzej