Scribd
Upload
63 views24 pages

CS5371 Theory of Computation: Lecture 9: Automata Theory VII (Pumping Lemma, Non-CFL)

This document provides an overview of automata theory and context-free languages (CFL). It introduces the pumping lemma for CFLs and applies it to show that certain languages are non-CFL. Specifically, it shows that the languages {anbncn}, {aibjck}, and {ww} are not CFL using proofs by contradiction with the pumping lemma. It also discusses that CFLs are closed under union, concatenation, and Kleene star but not complement or intersection. Finally, it reviews properties of CFLs and how the pumping lemma can be used to prove languages are non-CFL.

Uploaded by

Kamal Walia
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
63 views24 pages

CS5371 Theory of Computation: Lecture 9: Automata Theory VII (Pumping Lemma, Non-CFL)

This document provides an overview of automata theory and context-free languages (CFL). It introduces the pumping lemma for CFLs and applies it to show that certain languages are non-CFL. Specifically, it shows that the languages {anbncn}, {aibjck}, and {ww} are not CFL using proofs by contradiction with the pumping lemma. It also discusses that CFLs are closed under union, concatenation, and Kleene star but not complement or intersection. Finally, it reviews properties of CFLs and how the pumping lemma can be used to prove languages are non-CFL.

Uploaded by

Kamal Walia
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 24

CS5371

Theory of Computation
Lecture 9: Automata Theory VII
(Pumping Lemma, Non-CFL)
Objectives
•Introduce Pumping Lemma for CFL

•Apply Pumping Lemma to show that


some languages are non-CFL
Pumping Lemma for CFL
Theorem: If L is a CFL, then
there is a number p (pumping length)
where, if w is any string in L of length
at least p,
we can find u,v,x,y,z with w = uvxyz and
–for each i 0, uvixyiz is in L
–|vy| 0, and
–|vxy| p
Proof of Pumping Lemma
•Let b be the maximum branching factor in
the parse tree of any string in L
–that is, the right side of any rule has at most
b terminals and variables)
•We shall use p = b|V|+1 to prove the lemma
•Observation: What is the minimum
height of the parse tree for a string w
with length at least p?
Proof of Pumping Lemma (2)
•Height of the parse tree |V| + 1
 some path in tree |V|+2 nodes
•Only one such node can be a terminal
 at least |V|+1 variable on the path
•What does that mean?
Some variable appears at least twice
Proof of Pumping Lemma (3)
•Let R be a variable that
S
appears at least twice
•Then, the parse tree of
the string w looks R
something like:
R

w= u v x y z

So, uvixyiz is in L for any i 0 (why??)


uvixyiz is in L for any i 0
•Facts: R derives x, R derives vxy
•Since S derives uRz, and R S
derives x, S can derive uxz
R
S
x
R u z
R •Since S derives uvRyz
u v R y z
and R derives vxy,
S can derive uvvxyyz
v x y
Proof of Pumping Lemma (5)
•To complete the prove, we need to show
|vy| 0 and |vxy| p
•The current construction cannot, but
we can do so if we further restrict:
(1) parse tree is the smallest among all
that can generate the string w
(2) R is chosen from the lowest |V|+1
variables in the longest root-to-leaf
path
|vy| 0
•Suppose on the contrary that |vy| = 0
 Both v and y are empty strings
•Then in the parse tree, we replace
“
Subtree of R that generates vxy”
by “Subtree of R that generates x”
•Resulting parse tree will also generate
w (why?), but it has fewer nodes
 contradiction occurs
|vxy| p
• R is chosen from the lowest |V| + 1
variables in the longest root-to-leaf path
•Consider subtree of R that generates vxy
Its height is at most |V|+1 (why?)
 It has at most b|V|+1 leaves
 Thus, vxy has at most p characters
(as p = b|V|+1)
Recall: b = maximum branching factor
Non-CFL (example 1)
Theorem: The language
A = {anbncn | n 0}
is not a context-free language.

How to prove?
By contradiction, using pumping lemma
First thing: Assume that A is CFL
Proof (example 1)
•Let p be the pumping length
•Let w = apbpcp in A, and consider partition
w into any u,v,x,y,z such that w = uvxyz
•Two possible cases:
Case 1: Both v and y have only one type of char
Case 2: v or y has more than one type of char
•In both cases, uvvxyyz is not in A (why?)
•Thus, we find a string at least p long in A
that does not satisfy pumping lemma
 contradiction occurs
Non-CFL (example 2)
Theorem: The language
B = {aibjck | 0 i j k}
is not a context-free language.

How to prove?
By contradiction, using pumping lemma
First thing: Assume that B is CFL
Proof (example 2)
•Let p be the pumping length
•Let w = apbpcp in B, and consider partition
w into any u,v,x,y,z such that w = uvxyz
•Two possible cases:
Case 1: Both v and y have only one type of char
Case 2: v or y has more than one type of char
•We can see that for Case 2, uvvxyyz
cannot be in B
•How about Case 1?
Proof (example 2)
•Unfortunately, for Case 1, if v = b, y = c,
then the string uvvxyyz is always in B…
•So, how to get a contradiction??

•We divide Case 1 into two subcases:


Subcase 1.1: char a not appear in both v and y
Subcase 1.2: char a appears in v or y
Proof (example 2)
•For Subcase 1.1 (char a not appear in v and y),
uxz cannot be in B [why?]

•For Subcase 1.2 (char a appears in v or y),


uvvxyyz cannot be in B [why?]

•Thus, we find a string at least p long in B


that does not satisfy pumping lemma
 contradiction occurs
Non-CFL (example 3)
Theorem: The language
C = {ww | w in {0,1}*}
is not a context-free language.

How to prove?
By contradiction, use pumping lemma on
0p1p0p1p
Proof (example 3)
•When w = 0p1p0p1p = uvxyz, what can be
the corresponding vxy?
–Case 1: vxy appears in the first half
–Case 2: vxy appears in the second half
–Case 3: vxy includes the middle ‘10’
•For Cases 1 or 2, uvvxyyz not in C (why?)
•For Case 3, u must start with 0p, and z
must end with 1p (because |vxy| p and
vxy includes the middle ‘
10’)
 Then, uxz cannot be in C (why?)
CFL is closed under
all regular operations
•Union: We have seen that before
•Concatenation:
Let GA and GB be CFGs for two CFLs A
and B, using different sets of variables
Let SA and SB be their start variables
Combine the rules, add rule S  SA SB

•Star: Add rule S  S SA | 


CFL closed under complement?
•What is the complement of
A = {anbncn | n 0}? Assume = {a, b, c}

•The complement of A includes:


–strings containing ba, ca, or cb;
–strings aibjck with i j or j k
 the complement of A is a CFL (why??)

•As A is not a CFL, what can we conclude?


CFL closed under intersection?

•Is A = {anbncm | n,m 0} a CFL?


•Is B = {ambncn | n,m 0} a CFL?
•What is the intersection of A and B?
Is it a CFL?
•What can we conclude?
What we have learnt so far?
•PDA = CFG
–Prove by Construction
•Properties of CFG
- Ambiguous, Chomsky Normal Form
•Pumping Lemma
–Prove by Contradiction (using Parse Tree)
•Existence of non-CFL
Language Hierarchy
{ww}
Set of Languages (= set of “
set of strings”
) {0n1n2n}

{w with even |w|}

{w | w = wR}

{0x1y}
{0n1n}

Set of Regular Set of Context-


Language Free Language
Next Time
•Turing Machine
–A even more power computer

You might also like