Automata Theory

CS411-2004F-13
Unrestricted Grammars
David Galles
Department of Computer Science
University of San Francisco
13-0: Language Hierarchy
Regular
Languaes
Context Free
Languages
Regular Expressions
Finite Automata
Context-Free Grammars
Push-Down Automata
Recusively Enumerable
Languages
??
Turing Machines
13-1: CFG Review
G = (\. . 1. o)
\ = Set of symbols, both terminals & non-terminals
\ set of terminals (alphabet for the language
being described)
1 ((\ ) \

) Set of rules
o (\ ) Start symbol
13-2: Unrestricted Grammars
G = (\. . 1. o)
\ = Set of symbols, both terminals & non-terminals
\ set of terminals (alphabet for the language
being described)
1 (\

(\ )\

) Set of rules
o (\ ) Start symbol
13-3: Unrestricted Grammars
1 (\

(\ )\

) Set of rules
In an Unrestricted Grammar, the left-hand side of a
rule contains a string of terminals and
non-terminals (at least one of which must be a
non-terminal)
Rules are applied just like CFGs:
Find a substring that matches the LHS of some
rule
Replace with the RHS of the rule
13-4: Unrestricted Grammars
To generate a string with an Unrestricted
Grammar:
Start with the initial symbol
While the string contains at least one
non-terminal:
Find a substring that matches the LHS of
some rule
Replace that substring with the RHS of the
rule
13-5: Unrestricted Grammars
Example: Grammar for 1 = {c
n
/
n
c
n
: : 0}
First, generate (1C)

Next, non-deterministically rearrange string

Finally, convert to terminals ( c. 1 /,
etc.), ensuring that string was reordered to form
c

13-6: Unrestricted Grammars

Example: Grammar for 1 = {c
n
/
n
c
n
: : 0}
o 1Co
o 1
C
C C
1 1
C1 1C
C1
C
1
C
c
1
C
1
B
11
B
1
B
/
1
B
1
A
1
A
1
A
c
1
A
c
13-7: Unrestricted Grammars
o 1Co 1
A
//cc
1C1Co 1
A
c//cc
1C1Co 1
A
cc//cc
1C1Co cc//cc
11CCo
11CC1
C
11C1
C
c
111
C
cc
111
B
cc
11
B
/cc
1
B
//cc
13-8: Unrestricted Grammars
o 1Co 1111CCC1
C
1C1Co 111CC1
C
c
1C1C1Co 111C1
C
cc
1C1C1Co 1111
C
ccc
1C1C1Co 1111
B
ccc
1C1C1Co 111
B
/ccc
1C1C1Co 11
B
//ccc
1C1C1Co 1
B
///ccc
1C1C1Co 1
A
///ccc
11CC1Co 1
A
c///ccc
11C1CCo 1
A
cc///ccc
111CCCo 1
A
ccc///ccc ccc///ccc
13-9: Unrestricted Grammars
Example: Grammar for 1 = {uu : u c. /

}
13-10: Unrestricted Grammars
Example: Grammar for 1 = {uu : u c. /

}
Hints:
What if we created a string, and then
rearranged it (like (c/c)

c
n
/
n
c
n
)
13-11: Unrestricted Grammars
Example: Grammar for 1 = {uu : u c. /

}
Hints:
What if we created a string, and then
rearranged it (like (c/c)

c
n
/
n
c
n
)
What about trying uu
R
...
13-12: Unrestricted Grammars
1 = {uu : u c. /

}
o o

2
o

co

/o

1
o

c
2 A2 12 1 2
A A 1 1
1A A1 11 1 1
cA cc c1 c/
/A /c /1 //
13-13: Unrestricted Grammars
1
UG
is the set of languages that can be described
by an Unrestricted Grammar:
1
UG
= {1 : Unrestricted Grammar
G. 1[G] = 1}
Claim: 1
UG
= 1
re
To Prove:
Prove 1
UG
1
re
Prove 1
re
1
UG
13-14: 1
lG
1
:c
Given any Unrestricted Grammar G, we can create
a Turing Machine ` that semi-decides 1[G]
13-15: 1
lG
1
:c
Given any Unrestricted Grammar G, we can create
a Turing Machine ` that semi-decides 1[G]
Two tape machine:
One tape stores the input, unchanged
Second tape implements the derivation
Check to see if the derived string matches the
input, if so accept, if not run forever
13-16: 1
lG
1
:c
To implement the derivation on the second tape:
Write the initial symbol on the second tape
Non-deterministically move the read/write head
to somewhere on the tape
Non-deterministically decide which rule to apply
Scan the current position of the read/write
head, to make sure the LHS of the rule is at
that location
Remove the LHS of the rule from the tape, and
splice in the RHS
13-17: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
S
13-18: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
S
13-19: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
S ABC
13-20: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
S ABC
13-21: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
S ABCABC
13-22: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
S ABCABC
13-23: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
S AB C A BC
13-24: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
S AB C A BC
13-25: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
S AA C B BC
13-26: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
S AA C B BC
13-27: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
S AA B B CC
13-28: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
S AA B B CC
13-29: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
AA B B CCT
C
13-30: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
AA B B CCT
C
13-31: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
AA B B C c T
C
13-32: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
AA B B C c T
C
13-33: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
AA B B c c T
C
13-34: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
AA B B c c T
C
13-35: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
AA B B c c T
B
13-36: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
AA B B c c T
B
13-37: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
AA b B c c T
B
13-38: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
AA b B c c T
B
13-39: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
AA b b c c T
B
13-40: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
AA b b c c T
B
13-41: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
AA b b c c T
A
13-42: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
AA b b c c T
A
13-43: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
A a b b c c T
A
13-44: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
A a b b c c T
A
13-45: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
a a b b c c T
A
13-46: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
a a b b c c T
A
13-47: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
a a b b c c
13-48: 1
lG
1
:c
a
...
...
a b b c c
Input Tape
Work Tape
a a b b c c
13-49: 1
:c
1
lG
Given any Turing Machine ` that semi-decides
the language 1, we can create an Unrestricted
Grammar G such that 1[G] = 1
13-50: 1
:c
1
lG
Given any Turing Machine ` that semi-decides
the language 1, we can create an Unrestricted
Grammar G such that 1[G] = 1
Will assume that all Turing Machines accept in
the same conguration: (/. >)
Not a major restriction why?
13-51: 1
:c
1
lG
Given any Turing Machine ` that semi-decides
the language 1, we can create an Unrestricted
Grammar G such that 1[G] = 1
Will assume that all Turing Machines accept in
the same conguration: (/. >)
Not a major restriction why?
Add a tape erasing machine right before the
accepting state, that erases the tape, leaving
the read/write head at the beginning of the tape
13-52: 1
:c
1
lG
Given any Turing Machine ` that semi-decides
the language 1, we can create an Unrestricted
Grammar G such that 1[G] = 1
Grammar: Generates a string
Turing Machine: Works from string to accept
state
Two formalisms work in different directions
Simulating Turing Machine with a Grammar can be
difcult ..
13-53: 1
:c
1
lG
Two formalisms work in different directions
Simulate a Turing Machine in reverse!
Each partial derivation represents a
conguration
Each rule represents a backwards step in
Turing Machine computation
13-54: 1
:c
1
lG
Given a TM `, we create a Grammar G:
Language for G:
Everything in
M
Everything in 1
M
Start symbol o
Symbols > and
13-55: 1
:c
1
lG
Conguration (Q. >ncu) represented by the string:
>ncQu
For example, (Q. > c/c c) is represented by the
string > c/cQ c
13-56: 1
:c
1
lG
For each element in
M
of the form:
((Q
1
. c). (Q
2
. /))
Add the rule:
/Q
2
cQ
1
Remember, simulating backwards computation
13-57: 1
:c
1
lG
For each element in
M
of the form:
((Q
1
. c). (Q
2
. 1))
Add the rule:
Q
2
c cQ
1
13-58: 1
:c
1
lG
For each element in
M
of the form:
((Q
1
. ). (Q
2
. 1))
Add the rule
Q
2
Q
1

(undoing erasing extra blanks)

13-59: 1
:c
1
lG
For each element in
M
of the form:
((Q
1
. c). (Q
2
. 1))
Add the rule
c/Q
2
cQ
1
/
For all /
13-60: 1
:c
1
lG
For each element in
M
of the form:
((Q
1
. c). (Q
2
. 1))
Add the rule
c Q
2
cQ
1

(undoing moving to the right onto unused tape)

13-61: 1
:c
1
lG
Finally, add the rules:
o > /
> Q
s
c
c
13-62: 1
:c
1
lG
If the Turing machine can move from
>u to >/
Then the Grammar can transform
> Q
h
to > Q
s
u
Then, remove > Q
s
and to leave u
13-63: 1
:c
1
lG
Example:
Create a Turing Machine that accepts (c/)

,
halting in the conguration (/. >)
(assume conguration starts out as u)
13-64: 1
:c
1
lG
Example:
Create a Turing Machine that accepts (c/)

,
halting in the conguration (/. >)
R
a
R R
b
a
L yes
a,b
L yes
13-65: 1
:c
1
lG
c /

0
(
1
. ) (
1
. ) (
1
. )

1
(
2
. ) (
h
. )

2
(
3
. )

3
(
2
. ) (
4
. )

4
(
5
. ) (
5
. ) (
h
. )

5
(
4
. )
13-66: 1
:c
1
lG
((
0
. c). (
1
. ))
ccQ
1
cQ
0
c
c/Q
1
cQ
0
/
c Q
1
cQ
0

c Q
1
cQ
0

13-67: 1
:c
1
lG
((
0
. /). (
1
. ))
/cQ
1
/Q
0
c
//Q
1
/Q
0
/
/ Q
1
/Q
0

/ Q
1
/Q
0

13-68: 1
:c
1
lG
((
0
. ). (
1
. ))
cQ
1
Q
0
c
/Q
1
Q
0
/
Q
1
Q
0

Q
1
Q
0

13-69: 1
:c
1
lG
((
1
. c). (
2
. ))
ccQ
2
cQ
1
c
c/Q
2
cQ
1
/
c Q
2
cQ
1

c Q
2
cQ
1

13-70: 1
:c
1
lG
((
1
. ). (
h
. ))
/ Q
1
13-71: 1
:c
1
lG
((
2
. /). (
3
. ))
/cQ
3
/Q
2
c
//Q
3
/Q
2
/
/ Q
3
/Q
2

/ Q
3
/Q
2

13-72: 1
:c
1
lG
((
3
. c). (
4
. ))
ccQ
4
cQ
3
c
c/Q
4
cQ
3
/
c Q
4
cQ
3

c Q
4
cQ
3

13-73: 1
:c
1
lG
((
4
. c). (
5
. ))
Q
5
cQ
4
((
4
. /). (
5
. ))
Q
5
/Q
4
((
4
. ). (
h
. ))
/ Q
4
((
5
. ). (
4
. ))
Q
4
Q
5
13-74: 1
:c
1
lG
S h aQ
1
Q
0
a b Q
3
bQ
2

Q
0
bQ
1
Q
0
b b Q
3
bQ
2

Q
1
Q
0
aaQ
4
aQ
3
a
aaQ
1
aQ
0
a Q
1
Q
0
abQ
4
aQ
3
b
abQ
1
aQ
0
b aaQ
2
aQ
1
a a Q
4
aQ
3

a Q
1
aQ
0
abQ
2
aQ
1
b a Q
4
aQ
3

a Q
1
aQ
0
a Q
2
aQ
1
Q
5
aQ
4
baQ
1
bQ
0
a a Q
2
aQ
1
Q
5
bQ
4
bbQ
1
bQ
0
b h Q
1
h Q
4
b Q
1
bQ
0
baQ
3
bQ
2
a Q
4
Q
5
b Q
1
bQ
0
bbQ
3
bQ
2
b
13-75: 1
:c
1
lG
Generating c/c/
o > /
>/ >Q
4

> Q
4
> Q
5

> Q
5
> cQ
4

> cQ
4
> cQ
5

> cQ
5
> c/Q
4

> c/Q
4
> c/Q
5

> c/Q
5
> c/cQ
4

> c/cQ
4
> c/cQ
5

> c/cQ
5
> c/c/Q
4

13-76: 1
:c
1
lG
Generating c/c/
> c/c/Q
4
> c/c/Q
3

> c/c/ Q
3
> c/c/Q
2

> c/c/Q
2
> c/cQ
3
/
> c/cQ
3
/ > c/Q
2
c/
> c/Q
2
c/ > cQ
1
/c/
>cQ
1
/c/ >Q
0
c/c/
> Q
0
c/c/ c/c/>
c/c/ c/c/