is there any unambiguous grammar on alphabet={a,b} that can produce strings which have equal number of a and b (e.g. "aabb" , "baba" , "abba") ?
Asked
Active
Viewed 8,805 times
2 Answers
9
The problem with $S\to aSbS\mid bSaS\mid \varepsilon$ is that you're just making sure you match each $a$ with a $b$ (where we consider an $a$ and a $b$ to be matched iff they appeared during the same derivation step).
To ensure non-ambiguity, you must add a constraint on the matching to ensure that it is unique (while maintaining its existence). One way to do that is to make sure you match the first $a$ (resp. $b$) after your $b$ (resp. $a$) that hasn't been matched yet.
So you'd get something like $$ S\to aB S\mid bA S \mid \varepsilon\\ A \to a \mid b \square\\ B\to b \mid a\square$$
The idea is the following:
You want $S$ to generate words with as many $a$ as $b$s. At any points you can stop with the $S\to \varepsilon$. If you do add an $a$ with $S\to aBS$, then you need to add a $b$ later, and you put a $B$ to remind you of that, and then you continue with another $S$. The same things applies for $S\to bAS$.
If you have an $A$, it means that you are one $a$ short. If you read an $a$, everything is fine and you've got nothing else to do. But if you read a $b$, you are now two $a$s short. I left a $\square$ for you to fill to encode that.
$B$ works like $A$.
Solution :
You are two $a$s short so you are twice one $a$ short: $A \to a \mid b AA$. Similarly $B\to b\mid aBB$
For the proof, see this where $a,b,A,B$ are replaced with $0,1,O,I$ and your language is generated by $E$.
-1
The ambiguity of the above grammar can be resolved as
S->aBS/bAS/ε
A->a/bAA
B->b/aBB
