On closures of lexicographic star-free languages. E. Ochmański and K. Stawikowska

Similar documents
SYNTACTIC SEMIGROUP PROBLEM FOR THE SEMIGROUP REDUCTS OF AFFINE NEAR-SEMIRINGS OVER BRANDT SEMIGROUPS

CS 455/555: Finite automata

Aperiodic languages and generalizations

NOTES ON AUTOMATA. Date: April 29,

Closure Properties of Regular Languages. Union, Intersection, Difference, Concatenation, Kleene Closure, Reversal, Homomorphism, Inverse Homomorphism

Finite State Automata

A Weak Bisimulation for Weighted Automata

FORMAL LANGUAGES, AUTOMATA AND COMPUTABILITY

F. Blanchet-Sadri and F.D. Gaddis, "On a Product of Finite Monoids." Semigroup Forum, Vol. 57, 1998, pp DOI: 10.

Obtaining the syntactic monoid via duality

SEPARATING REGULAR LANGUAGES WITH FIRST-ORDER LOGIC

Great Theoretical Ideas in Computer Science. Lecture 4: Deterministic Finite Automaton (DFA), Part 2

Aperiodic languages p. 1/34. Aperiodic languages. Verimag, Grenoble

Büchi Automata and their closure properties. - Ajith S and Ankit Kumar

Axioms of Kleene Algebra

Finite Universes. L is a fixed-length language if it has length n for some

Duality and Automata Theory

Polynomial closure and unambiguous product

TECHNISCHE UNIVERSITÄT DRESDEN. Fakultät Informatik. Technische Berichte Technical Reports. Daniel Kirsten. TUD / FI 98 / 07 - Mai 1998

A Uniformization Theorem for Nested Word to Word Transductions

Probabilistic Aspects of Computer Science: Probabilistic Automata

Hierarchy among Automata on Linear Orderings

Bridges for concatenation hierarchies

COMP4141 Theory of Computation

Completeness of Star-Continuity

Deciding Whether a Regular Language is Generated by a Splicing System

Finite Automata and Regular languages

Mathematical Preliminaries. Sipser pages 1-28

CS 154, Lecture 3: DFA NFA, Regular Expressions

Tree languages defined in first-order logic with one quantifier alternation

Automata on linear orderings

Optimal Zielonka-Type Construction of Deterministic Asynchronous Automata

Theory of Computation

Algebra Meets Logic: The Case of Regular Languages (With Applications to Circuit Complexity) Denis Thérien, McGill University p.

CERNY CONJECTURE FOR DFA ACCEPTING STAR-FREE LANGUAGES

Equivalence of DFAs and NFAs

Semi-simple Splicing Systems

MATH 433 Applied Algebra Lecture 22: Semigroups. Rings.

FORMAL LANGUAGES, AUTOMATA AND COMPUTABILITY

CS 154, Lecture 2: Finite Automata, Closure Properties Nondeterminism,

THEORY OF COMPUTATION (AUBER) EXAM CRIB SHEET

Constructive Formalization of Regular Languages

Equational Theory of Kleene Algebra

Automata Theory. Lecture on Discussion Course of CS120. Runzhe SJTU ACM CLASS

On Properties and State Complexity of Deterministic State-Partition Automata

Asynchronous cellular automata for pomsets. 2, place Jussieu. F Paris Cedex 05. Abstract

DM17. Beregnelighed. Jacob Aae Mikkelsen

Theory of Computation (I) Yijia Chen Fudan University

CMPSCI 250: Introduction to Computation. Lecture #29: Proving Regular Language Identities David Mix Barrington 6 April 2012

2. Syntactic Congruences and Monoids

CS 154. Finite Automata, Nondeterminism, Regular Expressions

Introduction to Kleene Algebra Lecture 13 CS786 Spring 2004 March 15, 2004

Weighted automata and weighted logics

Algebraic Approach to Automata Theory

Unit 6. Non Regular Languages The Pumping Lemma. Reading: Sipser, chapter 1

Uses of finite automata

An algebraic characterization of unary two-way transducers

Languages, logics and automata

Congruence Boolean Lifting Property

We define the multi-step transition function T : S Σ S as follows. 1. For any s S, T (s,λ) = s. 2. For any s S, x Σ and a Σ,

Equational Logic. Chapter Syntax Terms and Term Algebras

Equivalence of Regular Expressions and FSMs

Automata extended to nominal sets

Local LTL with past constants is expressively complete for Mazurkiewicz traces

Theory of Computation (II) Yijia Chen Fudan University

On the Accepting Power of 2-Tape Büchi Automata

Lecture Notes: Selected Topics in Discrete Structures. Ulf Nilsson

UNIT-II. NONDETERMINISTIC FINITE AUTOMATA WITH ε TRANSITIONS: SIGNIFICANCE. Use of ε-transitions. s t a r t. ε r. e g u l a r

On Recognizable Languages of Infinite Pictures

Warshall s algorithm

CSE 105 THEORY OF COMPUTATION. Spring 2018 review class

Jumping Finite Automata

Finite Automata and Regular Languages

Inf2A: Converting from NFAs to DFAs and Closure Properties

Pumping for Ordinal-Automatic Structures *

Finite n-tape automata over possibly infinite alphabets: extending a Theorem of Eilenberg et al.

The submonoid and rational subset membership problems for graph groups

Duality in Logic. Duality in Logic. Lecture 2. Mai Gehrke. Université Paris 7 and CNRS. {ε} A ((ab) (ba) ) (ab) + (ba) +

Computational Models - Lecture 5 1

Properties of Languages with Catenation and

CHAPTER 1. Relations. 1. Relations and Their Properties. Discussion

CS 275 Automata and Formal Language Theory

On the Regularity of Binoid Languages:

Foundations of Informatics: a Bridging Course

SOLVABILITY OF EQUATIONS IN GRAPH GROUPS IS DECIDABLE

Properties of Context-Free Languages. Closure Properties Decision Properties

On Recognizable Languages of Infinite Pictures

Definition of Büchi Automata

Automata Theory for Presburger Arithmetic Logic

CSE 105 THEORY OF COMPUTATION

languages by semifilter-congruences

Varieties Generated by Certain Models of Reversible Finite Automata

L is finite or cofinite}, A + k+1 = { L A + L is a boolean combination of languages of the form L 1 L n (n 1) with L 1, L n A +

Final exam study sheet for CS3719 Turing machines and decidability.

Formal Models in NLP

Closure Under Reversal of Languages over Infinite Alphabets

Finite Automata Theory and Formal Languages TMV027/DIT321 LP4 2017

Languages and monoids with disjunctive identity

group Jean-Eric Pin and Christophe Reutenauer

6.8 The Post Correspondence Problem

Transcription:

On closures of lexicographic star-free languages E. Ochmański and K. Stawikowska Preprint No 7/2005 Version 1, posted on April 19, 2005

On closures of lexicographic star-free languages Edward Ochma ski and Krystyna Stawikowska Faculty of Mathematics and Computer Science, Nicolaus Copernicus University, Toru {edoch,entropia}@mat.uni.torun.pl Abstract. Muscholl and Petersen showed that, in the case of transitive dependencies, closures of star-free lexicographic languages are star-free or non-regular. It implies that, in the same case of transitive dependencies, closures of star-free lexicographic languages are star-free. In this paper, it is shown to be true also in the case of transitive independencies. Main result is even more general, but the general question, if closures of star-free lexicographic languages are star-free in any case, remains open. 1 Introduction Trace theory, as a formal tool for description of concurrent behaviours, was proposed by Mazurkiewicz [2]. The approach is based on the notion of independency relation, expressing concurrent execution of actions. One of the most useful operations on languages, related to independency, is the closure operation, introduced by Szíjártó [8] in the initial period of trace theory. Theory of star-free languages started with the fundamental Schützenberger [7] Theorem (star-free = aperiodic in free monoids). Guaiana/Restivo/Salemi [1] combined the both subjects, proving Schützenberger Theorem for arbitrary trace monoids. The main inspiration for the present paper descends from Muscholl/Petersen [4], where the closure operation in the context of star-free languages was studied. Lexicographic words play an important role in the theory of recognizable trace languages, because Closures of regular sets of lexicographic words are again regular (EO [5]). Results of Muscholl/Petersen [4] arouse a suspicion that Closures of star-free sets of lexicographic words are again star-free. It is true for trace monoids with transitive dependencies (cartesian products of free monoids), by results of [4] and [5]. We prove in this paper that it is also true for trace monoids with transitive independencies (free products of commutative monoids). For this aim, we define a subclass of star-free languages (called LSF-languages) and show that each star-free language of lexicographic words is an LSF-language. The proof consists in a detailed analysis of minimal deterministic automata for sets of lexicographic words. 1

2 Preliminaries In this section, we recall very briefly basic notions and known results, needed here. Finally, we state the main question of the paper. 2.1 Basic Notions An alphabet A is assumed to be finite. The set A* with concatenation as the product operation form the free monoid; subsets of A* are called (word) languages. Let I A A be a symmetric and irreflexive relation on A, called independency, its complement D=A A I is named dependency. The couple (A,I) or (A,D) is said to be a concurrent alphabet. Given a concurrent alphabet (A,I), the trace monoid A*/I is the quotient of the free monoid A* by the least congruence on A* containing the relation {ab=ba aib}. Members of A*/I are called traces, and sets of traces (i.e. subsets of A*/I) are called trace languages. Clearly, a trace monoid A*/I is free iff I=. Given a monoid M, the complement of a subset X M will be denoted by, i.e. X =M X. Let (A,I) be a concurrent alphabet. Any word w A* induces a trace [w] A*/I the congruency class of w, any word language L A* induces a trace language [L] = {[w] w L} the set of all traces induced by members of L. Given a trace language T A*/I, the flattening of T is the word language T = {w A* [w] T} the union of traces in T, viewed as subsets of A*. Given a word language L A*, the closure of L is the word language L =[L]. A word language L is said to be closed (w.r.t. I) iff L= L. Given a trace monoid M=A*/I and a trace language T M, the following notions are well-known: atomic trace languages (atoms, for short):, [ε], all [a] for a A; syntactic congruence T M M of T: x T y iff ( r,s M) rxs T rys T; syntactic monoid of T: the quotient monoid M/ T. A trace language T M=A*/I is said to be: rational iff it is built from atoms with union, product and star; recognizable iff its syntactic monoid is finite; star-free iff it is built from atoms with union, product and complement; aperiodic iff its syntactic monoid is finite and aperiodic ( n) ( x) x n =x n+1. Classes of languages, defined above, will be denoted by RAT(M), REC(M), SF(M) and AP(M), respectively. The argument will be possibly omitted, if it will not lead to a confusion. As RAT(A*)=REC(A*) in finitely generated free monoids (Kleene Theorem), the class RAT(A*)=REC(A*) will be uniformly denoted by REG(A*) in that case. By definitions, AP(M) REC(M) for any monoid M. Moreover, if M is a trace monoid, there hold the inclusions SF(M) REC(M) RAT(M). 2

2.2 Roots and influences Let us remind some fundamental results, utilized in the paper. Theorem 2.1 (Schützenberger [7]). SF(A*)=AP(A*). Theorem 2.2 (Guaiana/Restivo/Salemi [1]). Let M=A*/I be a trace monoid and T M. T SF(M) T AP(M) T AP(A*) T SF(A*) Corollary 2.3. The family of closed SF-languages is the least family containing atoms and closed under union, complement and closed product (where X Y= XY ). Corollary 2.4. Let X,Y A*. If X,Y SF, then XY = X Y SF. Theorem 2.5 (Muscholl/Petersen [4]). Let (A,D) be a concurrent alphabet with transitive dependency D, and let L A*. If L SF then L SF or L REG. Example 2.6 (Muscholl/Petersen [4]). If D is not transitive, then Theorem 2.5 does not hold, as shown by the set L=(abcbac)* for D: a c b. 2.3 Lexicographic languages and their closures A concurrent alphabet, equipped with a strict order on the alphabet, will be called an oriented concurrent alphabet and denoted by (A,<,I). Such an alphabet is said to be transitively oriented iff the relation < I is transitive. Any strict order on A induces the well-known lexicographic order on A*. A word w A* is said to be lexicographic (w.r.t. < and I) iff it is lexicographically first in its closure w A*. The set of all lexicographic words will be denoted by LEX (assuming < and I are unambiguously fixed). Notice that, for each oriented concurrent alphabet, LEX is star-free, as LEX = A* {A*bI a *aa* aib a<b}, where I a ={c A aic}. Property 2.7. ( x,y,z A*) xyz LEX y LEX. Given an oriented concurrent alphabet (A,<,I), the lexicographic representation of a trace language T A*/I is defined as the word language Lex(T)=T LEX. Theorem 2.8 (EO [5]). Let M=A*/I be a trace monoid. trace formulation T REC(M) Lex(T) REG(A*) word formulation L REG(A*) L LEX REG(A*) 3

Theorem 2.8 arouses a question for star-free languages: Question 2.9. Let (A,<,I) be an oriented concurrent alphabet. Is it true that trace formulation T SF(A*/I)? Lex(T) SF(A*), word formulation L SF(A*)? L LEX SF(A*). The question is supported by the result and example of Muscholl/Petersen [4], as Theorems 2.5 and 2.8 yield the positive answer in the case of transitive dependency (as SF(A*) REG(A*) and L LEX = L ), and moreover, the language of Example 2.6 does not work as counterexample for our question, because it is not included in LEX. Remark that the implication is true, because the family SF is closed under intersection. Thus the crucial question may be formulated as follows: Question 2.10. Does L SF L LEX imply L SF? 3 LSF-languages Let us define two operations on languages, related to LEX. lexicographic product X Y = XY if XY LEX and undefined otherwise; lexicographic complement X = LEX X. Let LSF be the class of languages built from atoms with union, lexicographic product and lexicographic complement. Any expression built from atoms with, (whenever defined) and is called an LSF-expression. Fact 3.1. If L LSF then L SF and L LEX Proof. Obvious, as atoms and LEX are star-free and included in LEX. We will see, in the next section, that the converse of Fact 3.1 is true in the transitively oriented case, as a consequence of Lemmas 4.6 and 4.7. We do not know, if it is true in general. Lemma 3.2. If L LSF then L SF. Proof. Structural induction on LSF-expressions: For atoms obvious, for because X Y = X Y, for from Corollary 2.4, for because X = A* X if X LEX. 4

4 Transitively oriented case For this section, we assume (A,<,I) with transitive < I. In this case the following characterization of LEX holds: Proposition 4.1 (Métivier/EO [3]). If (A,<,I) is transitively oriented, then LEX = {w A* ( a,b A) ( u,v A*) w=uabv adb a<b}. Let us denote, for y A, ylex = LEX ya* lexicographic words started with y LEX y = LEX A*y lexicographic words ended in y Lemma 4.2. If (a 1 <...< a n ;I) is transitively oriented, then ( y A) y LEX LSF and LEX y LSF Proof. Notice that, as a consequence of Proposition 4.1, we have (1) ( y A) y LEX = y (LEX { x LEX x<y xiy}), (2) ( y A) LEX y = (LEX {LEX z y<z yiz}) y. Clearly, (1 ) a1 LEX = a 1 LEX LSF and (2 ) LEX an = LEX a n LSF. Now, inductively from a 1 to a n, starting with (1 ), we get (1 ) ( y A) y LEX LSF, and from a n to a 1, starting with (2 ), we get (2 ) ( y A) LEX y LSF. 4.1 Automata for LEXes Automaton is a quadruple (A,Q,s 0,F), where A is an alphabet, Q is a set of states, s 0 Q is an initial state and F Q is a set of final states; states are partial functions q:a Q; domain of q will be denoted by dom(q). Whenever F=Q (any state is final), we write such automaton as a triple (A,Q,s 0 ); it is the case of automata for LEXes. Proposition 4.1 gives reasons, in the case of transitively oriented alphabets, for the following construction of minimal deterministic automaton A LEX for the set LEX. Construction 4.3. M.d.a. for LEX transitively oriented case Given a transitively oriented concurrent alphabet (a 1 <... < a m ;D), the automaton is built up inductively, as follows. A 1 = A LEX (a 1 ;D): ({a 1 },{q 1 }, q 1 ), where q 1 (a 1 )=q 1 Given the automaton A n = A LEX (a 1 <... < a n ;D) = ({a 1,...,a n }, Q n, q 1 ) for n<m (set Q n =k), we build A n+1 = A LEX (a 1 <... < a n < a n +1 ;D) = ({a 1,...,a n,a n+1 }, Q n+1, q 1 ): 5

1 Q n+1 =Q n {q}, where q=q i Q n if ( x dom(q i )) xda n+1 and ( j<i) ( x dom(q i )) xia n+1 in words, q i is the first state of Q n such that all members of its domain depend on a n+1. If such a state does not exist, q=q k+1 a new state is added. 2 We extend domains of states in Q n : ( i k) q i (a n+1 )=q (the state found or added in 1 ). 3 If a new state was added in 1, we define its activity: dom(q k+1 )={x {a 1,...,a n,a n+1 } xda n+1 }; q k+1 (x)=q i iff ( y dom(q i )) ydx x<y and ( j<i) ( y dom(q i )) yix y<x in words, q i is the first state of Q n+1 such that each member of its domain depends on x or follows x (in the alphabetic order). Example 4.4. For the concurrent alphabet (A,D): a c b d, with the order a<b<c<d, the construction of m.d.a. for LEX proceeds as follows (from left to right: A 1, A 2, A 3 and A 4 ): a 1 a 1 a 1 c a 1 c d b c b c b 3 d b d b 2 b 2 b 2 Property 4.5. Let (A,<,I) be transitively oriented and A LEX = (A,Q,q 1 ). Then, for each a A, there is exactly one state q a Q such that ( p,r Q) p(a)=r r=q a i.e. all a-labelled arcs in A LEX aim at a common state q a. Proof. Directly from Construction 4.3. Given an automaton A LEX = (A,Q,q 1 ), by L(p,r) we denote, for p,r Q, the language recognizable by the automaton (A,Q,p,r), i.e. the language given by all paths in A LEX from p (as initial state) to r (as final state). And by L(p,Q) we denote {L(p,r) r Q}. Lemma 4.6. Let (A,<,I) be transitively oriented and A LEX = (A,Q,q 1 ). Then ( p,r Q) L(p,r) LSF. Proof. We have L(q 1,r)={LEX y ( q Q) q(y)=r} ( {ε} if r=q 1 ), by Property 4.5, and L(p,Q)={ y LEX y dom(p)} {ε}, by Properties 2.7 and 4.5. Hence, by Lemma 4.2, L(q 1,r) and L(p,Q) are in LSF. Now observe that L(p,r) {ε} = L(p,Q) L(q 1,r) {ε}, by Properties 2.7 and 4.5. As LSF is closed under boolean operations, L(p,r) LSF. 6

4.2 Main result The following lemma holds for arbitrary oriented alphabets. Lemma 4.7. Let (A,<,I) be an oriented concurrent alphabet and A LEX = (A,Q,q 1 ). If any L(p,r) LSF, then L SF L LEX L LSF. Proof. First we prove, under the hypothesis of the lemma, the following claim. Claim: ( L SF)( p,r Q) L L(p,r) LSF. Proof. Structural induction on SF-expression. For atoms obvious. Induction step: (X Y) L(p,r) = (X L(p,r)) (Y L(p,r)) LSF; XY L(p,r) = {(X L(p,q))(Y L(q,r)) q Q} LSF; X L(p,r) = X L(p,r) = L(p,r) (X L(p,r)) LSF. End of Claim Now observe that LEX={L(q 1,r) r Q}, thus L LEX={L L(q 1,r) r Q}. As all L L(q 1,r) LSF (by Claim), its union is in LSF, too. Remark that Lemmas 4.6 and 4.7 yield the converse of Fact 3.1 in the case of transitively oriented concurrent alphabets. And now we can prove: Proposition 4.8. Let (A,<,I) be a transitively oriented concurrent alphabet. If L SF and L LEX, then L SF. Proof. Any L(p,r) LSF, by Lemma 4,6. Then L LEX=L LSF, by Lemma 4.7. And Lemma 3.2 ends the proof. 5 Conclusions Let us recall Theorem 2.5 and notice that, with a support of Theorems 2.8 and 2.2, it yields Proposition 5.1. Let (A,I) be a concurrent alphabet with transitive D=A* A* I, and let T A*/I be a trace language. Then (i) T SF iff (ii) ( <) Lex < (T) SF iff (iii) ( <) Lex < (T) SF Proof. (iii) (ii) is obvious. (ii) (i): Let L=Lex < (T); by Theorem 2.8, L REG; then by Theorem 2.5, T= L SF; and finally, by Theorem 2.2, T SF. (i) (iii): by Theorem 2.2, T SF, thus Lex < (T)= T LEX SF, since LEX SF. Remark that only (ii) (i) uses the assumption that D is transitive (Theorem 2.5). 7

Let us look at concurrent alphabets with transitive I. Remind that such trace monoids constitute an important class of trace monoids. Namely, it is just the class of trace monoids with decidable recognizability problem (Sakarovitch [6]) and SF-problem (Muscholl/Petersen [4]). Results of the present paper show that thesis of Proposition 5.1 is true also under the hypothesis of transitive I. Proposition 5.2. Let (A,I) be a concurrent alphabet with transitive I, and let T A*/I be a trace language. Then (i) T SF iff (ii) ( <) Lex < (T) SF iff (iii) ( <) Lex < (T) SF Proof. From Proposition 4.8, as in this case (A,<,I) is transitively oriented for any <. The question, if Propositions 5.1 and 5.2 hold unconditionally (for arbitrary concurrent alphabets), remains open. Notice that there are concurrent alphabets with non-transitive D and without transitive orientations, for example the pentagon (A,D): References 1. G. Guaiana, A. Restivo, S. Salemi: Star-free trace languages. Theoretical Computer Science 97, pp. 301-311, 1992. 2. A. Mazurkiewicz: Concurrent program schemes and their interpretations. Report DAIMI-PB-78, Aarhus University, 1977. 3. Y. Métivier, E. Ochma ski: On lexicographic semi-commutations. Information Processing Letters 26, pp. 55-59, 1987. 4. A. Muscholl, H. Petersen: A note on the commutative closure of star-free languages. Information Processing Letters 57, pp. 71-74, 1996. 5. E. Ochma ski: Regular behaviour of concurrent systems. Bulletin of EATCS 27, pp. 56-67, 1985. 6. J. Sakarovitch: The last decision problem for rational trace languages. Proc. of LATIN 92, LNCS 583, pp. 460-473. Springer, 1992. 7. M.P. Schützenberger: On finite monoids having only trivial subgroups. Information and Control 8, pp. 190-194, 1965. 8. M. Szíjártó: A classification and closure properties of languages for describing concurrent system behaviours. Fundamenta Informaticae 4, pp. 531-549, 1981. 8