Normal Forms for Context-Free Grammars

CS 3240 – Chapter 6 Normal Forms for Context-Free Grammars

Topics • 6.1: Simplifying Grammars • Substitution • Removing useless variables • Removing λ • Removing unit productions • 6.2: Normal Forms • Chomsky Normal (CNF) • 6.3: A Membership Algorithm • CYK Algorithm • An example of bottom-up parsing CS 3240 - Normal Forms for Context-Free Languages

Simplifying Context-Free Grammars • Variables in CFGs can often be eliminated • If they are not recursive, you can substitute their rules throughout • See next slide CS 3240 - Normal Forms for Context-Free Languages

Detecting Cycles and Useless Variables • A variable is useless if: • It can’t be reached from the start state, or • It never leads to a terminal string • Due to endless recursion • Both problems can be detected by a dependency graph • See next slide CS 3240 - Normal Forms for Context-Free Languages

Useless Variables S ➞ aSb | A | λ A ➞ aA A is useless (non-terminating): S ➞ aSb | λ CS 3240 - Normal Forms for Context-Free Languages

Useless Variables S ➞ A A ➞ aA | λ B ➞ bA B is useless (non-reachable): S ➞ A A ➞ aA | λ CS 3240 - Normal Forms for Context-Free Languages

Exercise Simplify the following: S ➞ aS | A | C A ➞ a B ➞ aa C ➞ aCb CS 3240 - Normal Forms for Context-Free Languages

A Dependency Graph CS 3240 - Normal Forms for Context-Free Languages

Nullable Variables • Any variable that can eventually terminate in the empty string is said to be nullable • Note: a variable may be indirectly nullable • In general: if A ➞ V1V2…Vn, and all the Vi are nullable, then A is also nullable • See Theorem 6.3 CS 3240 - Normal Forms for Context-Free Languages

Finding Nullable Variables Consider the following grammar: S ➞ a | Xb | aYa X ➞ Y | λ Y ➞ b | X Which variables are nullable? How can we substitute the effect of λ before removing it? CS 3240 - Normal Forms for Context-Free Languages

Removing λ • First find all nullable variables • Then substitute (A + λ) for every nullable variable A, and expand • Then remove λ everywhere from the grammar • What’s left is equivalent to the original grammar • except the empty string may be lost • we won’t worry about that CS 3240 - Normal Forms for Context-Free Languages

Removing λExample Consider the following grammar (again): S ➞ a | Xb | aYa X ➞ Y | λ Y ➞ b | X How can we substitute the effect of λ before removing it? CS 3240 - Normal Forms for Context-Free Languages

Removing Unit Productions • Unit Productions often occur in chains • A ➞ B ➞ C • Must maintain the effect of BandC when substituting for A throughout • Procedure: • Find all unit chains • Rebuild grammar by: • Keeping all non-unit productions • Keeping only the effect of all unit productions/chains CS 3240 - Normal Forms for Context-Free Languages

Removing Unit ProductionsExample S ➞ A | bb A ➞ B | b B ➞ S | a Note that S ⇒* {A,B}, A⇒* {B,S}, B ⇒* {S,A} Giving: S ➞ bb | b | a // Added non-unit part of A and B A ➞ b| a | bb // Added non-unit part of B and S B ➞ a | bb | b// Added non-unit part of S and A CS 3240 - Normal Forms for Context-Free Languages

Unit Rule Removal Procedure • 1) Determine all variables reachable by unit rules for each variable • 2) Keep all non-unit rules • 3) Substitute non-unit rules in place of each variable reachable by unit productions CS 3240 - Normal Forms for Context-Free Languages

Why Remove Nulls First? S ➞ aA A ➞ BB B ➞ aBb | λ Now remove nulls and see what happens…. (Also see the solution for #15 in 6.1) CS 3240 - Normal Forms for Context-Free Languages

Remove Nulls and Units S ➞ AB A ➞ B B ➞ aB | BB | λ CS 3240 - Normal Forms for Context-Free Languages

Simplification Summary • Do things in the following recommended order: • Remove nulls • Remove unit productions • Remove useless variables • Simplify by substitution as desired CS 3240 - Normal Forms for Context-Free Languages

Chomsky Normal FormSection 6.2 • Very important for our purposes • All CNF rules are of one of the following two forms: • A ➞ c (a single terminal) • A ➞ XY (exactly two variables) • Must begin the transformation aftersimplifying the grammar (removing λ, all unit productions, useless variables, etc.) CS 3240 - Normal Forms for Context-Free Languages

Exercise Convert the following grammar to CNF: S ➞ abAB A ➞ bAB | λ B ➞ BAa | A | λ CS 3240 - Normal Forms for Context-Free Languages

Greibach Normal Form • Single terminal character followed by zero or more variables (cV*, c ∈ Σ ) • V → a • V → aBCD… • λ allowed only in S → λ • Sometimes need to make up new variable names CS 3240 - Normal Forms for Context-Free Languages

Greibach Example 1 S → AB A → aA | bB | b B → b Substitute for A in first rule (i.e., add B to each rule for A): S → aAB | bBB | bB The other rules are okay CS 3240 - Normal Forms for Context-Free Languages

Greibach Example 2 S → abSb |aa Add rules to generate a and b: S → aBSB |aA A → a B → b CS 3240 - Normal Forms for Context-Free Languages

The CF Membership ProblemSection 6.3 • The “parsing” problem • How do we know if a string is generated by a given grammar? • Bottom-up parsing (CYK Algorithm) • An Example of Dynamic Programming • Requires Chomsky Normal Form (CNF) • Start by considering A ➞ c rules • Build up the parse tree from there CS 3240 - Normal Forms for Context-Free Languages

A Parsing Example S ➞ XY X ➞ XA | a | b Y ➞ AY | a A ➞ a Does this grammar generate “baaa”? CS 3240 - Normal Forms for Context-Free Languages

Parse Trees for “baaa” CNF yields binary trees.(Can you find a third parse tree?) CS 3240 - Normal Forms for Context-Free Languages

Parsing “baaa” S ➞ XY X ➞ XA | a | b Y ➞ AY | a A ➞ a Stage 1: b ⇐ X a ⇐ X, Y, A Stage 2: ba ⇐ X(X,Y,A) = XX, XY, XA ⇐ S, X aa ⇐ (X,Y,A)(X,Y,A) = XX, XY, XA, YX, YY, YA, AX, AY, AA ⇐ S, X, Y Stage 3: baa ⇐ ba a ⇐ (S,X)(X,Y,A) = SX, SY, SA, XX, XY, XA = S, X baa ⇐ baa ⇐ X(S,X,Y) = XS, XX, XY = S aaa ⇐ aa a ⇐ (S,X,Y)(X,Y,A) = _________ aaa ⇐ a aa ⇐ (X,Y,A)(S,X,Y) = _________ Stage 4: (you finish…) baaa: baaa ⇐ __________ baaa ⇐ __________ baa a ⇐ __________ CS 3240 - Normal Forms for Context-Free Languages

CYK AlgorithmTabular Form – Stage 1 CS 3240 - Normal Forms for Context-Free Languages

Exercise Does the following grammar generate “abbaab”? S ➞ SAB | λ A ➞ aA | λ B ➞ bB | λ CS 3240 - Normal Forms for Context-Free Languages

Normal Forms for Context-Free Grammars