2

Are there any examples of context-free grammars for which we simply do not know whether they are ambiguous? By examples I mean an actual specification for the grammar, not some kind of non-constructive existence proof with a counting argument or whatnot. To make things more clear, I should be able to write a parser for this grammar from the specification (this should exclude grammars that are somehow defined via constructions involving other undecidable problems).

I am not a specialist so I hope the previous paragraph makes what I mean clear enough. Otherwise please do let me know if you think I am not being clear enough.

Another point: coming up with a random grammar specification on the spot and claiming its ambiguity is unknown because no one has tried to figure it out does not count. I guess I am really looking for known open problems in this area.

Tob Ernack
  • 141
  • 4
  • 1
    You have succumbed to a common misconception: undecidability does not imply that a single instance must be "unclear". See also here. Duplicate? – Raphael Jan 09 '18 at 23:05
  • I see. So it turns out my question is equivalent to asking for specific examples of Turing machines whose halting we do not know? – Tob Ernack Jan 09 '18 at 23:11
  • I suppose that makes sense because of the reductions to the Post correspondence problem which itself reduces to the halting problem. – Tob Ernack Jan 09 '18 at 23:16
  • And for the fallacy you describe in the linked question: in your answer you state that one of the algorithms that always outputs Yes or always outputs No will correctly decide for a specific instance of the problem. So is your point that the problem is decidable because such an algorithm exists, even if we don't know which one is actually correct? – Tob Ernack Jan 09 '18 at 23:26
  • I'm not saying it's logically equivalent, I'm saying the wrong misunderstanding fuels both questions. – Raphael Jan 09 '18 at 23:31
  • "So is your point that the problem is decidable because such an algorithm exists, even if we don't know which one is actually correct?" -- Yes, that is my point. And the definition of "decidable". – Raphael Jan 09 '18 at 23:31
  • 1
    Thank you for your reply. So when I say "we cannot even prove that the ambiguity problem is undecidable for this particular grammar", I am wrong because we can, and the answer is "it is decidable". In that case, I will retract this claim, but my question still stands: I am interested in particular cases where CFG ambiguity has not been determined, as far as public research knowledge goes. – Tob Ernack Jan 09 '18 at 23:50
  • Now I am starting to wonder if you could construct such an example by the same methods used in the question about Turing machines (choose a conjecture like Collatz, etc.) and translate somehow into a CFG. You did not specify whether the question is logically equivalent to that of halting for a specific Turing machine, so perhaps this construction does not actually work. – Tob Ernack Jan 09 '18 at 23:53
  • No, I wouldn't think that such grammars and TMs are "equivalent", beyond that some (in general) undecidable property is known for them. That said, similar techniques may work. However, I don't think you'll find "closed" definitions of grammars that have the desired property. Tricks like "include this rule if is true" are probably not what you're looking for. – Raphael Jan 10 '18 at 08:04

0 Answers0