More advanced format

[nordlow]

I would like to enable pattern matching in formatting. Something like

(format EnglishLanguage
   (holdsDuring
      (PastFn ?TIMEPOINT)
      ?FORMULA)
   (stringConcatenation ?FORMULA "before" ?TIMEPOINT))

(format EnglishLanguage
   (holdsDuring
      (FutureFn ?TIMEPOINT)
      ?FORMULA)
   (stringConcatenation ?FORMULA "since" ?TIMEPOINT))

Is this proposal a good idea?

[arademaker]

I see two ways to improve the language generation from KIF expressions:

1. Follow the ideas from [1], review what can be improved and how.

2. I don't know much about the Java code from https://github.com/ontologyportal/sigmakee. In our https://github.com/own-pt/cl-krr some language generation could be implemented as Lisp macros with reasonable simplicity following the ideas of SigmaKEE and improving with your suggestion. PR are welcome.

(1) is more robust, for sure.

[1] R. Enache, “Reasoning and Language Generation in the SUMO Ontology,” University of Gothenburg, 2010.

[apease]

I don't recall enough about Ramona's proposal to say. I recall pointing out a number of issues with her understanding of SUMO. But definitely a more comprehensive and sophisticated approach to NLG in SUMO is needed. A good start would be using prepositions properly with Processes, so that instead of "the agent of Walking is John-1" we could have "John-1 walks". The current NLG system was developed to handle languages like English that are only lightly inflected, and it can't handle Slavic or Nordic languages (or even romance languages) that have much richer forms of agreement. But I'd caution against taking a piecemeal approach doing just a language feature at a time and instead try to collect a set of example SUMO axioms, then express them in a colloquial way in several target languages, and then develop the mechanisms.