Re: Link grammars in Programming Language Development?

"Ray" <bear@sonic.net> wrote in message news:10-02-060@comp.compilers...
> There has been a lot of research on Link grammars for natural language
> processing. They do pretty well, and link failures (grammar doesn't
> match input) seem well-correlated to very specific error conditions.
>
> So it occurred to me to wonder whether anyone's using them for
> programming language development? Or, specifically, whether there
> are any tools similar in spirit to Lex/Yacc/etc that use link
> grammars rather than production grammars?


I am not personally aware of any languages using them.


the issue WRT link grammars is that they would likely have a notable impact
on the language syntax, and so to use this sort of grammar in the design of
a programming language would, likely, involve having to re-imagine many of
the basic syntactic forms. this could also potentially impact the semantics
as well.


I have doubts it would be a good match for existing common programming
languages (such as the C-family, for example), but then again, not having
much experience with the grammars I could be wrong.




as for tools, one link that came up was this:
http://www.abisource.com/projects/link-grammar/


although I have had no real personal experience with any of this.




as for myself, actually, I have not really seen or worked with these
grammars.


I had done some things related to parsing English, but this had approached
things from a different stance:
I had essentially tried to "mechanize" the English, mostly by limiting the
number of valid syntactic constructions, enforcing a "one word, one grammar
type" rule, and by attempting to use a recursive descent parser.


however, what became a problem here was that I couldn't come up with a good
semantic model, and didn't have a particular use-case, which sort of limited
the point of parsing it for my uses.


some of my other uses of natural language typically didn't require full
parsers.

Return to the comp.compilers page.
Search the comp.compilers archives again.