DocumentCode
3112551
Title
A Contraction Method to Decide MSO Theories of Deterministic Trees
Author
Montanari, Angelo ; Puppis, Gabriele
Author_Institution
Univ. di Udine, Udine
fYear
2007
fDate
10-14 July 2007
Firstpage
141
Lastpage
150
Abstract
In this paper we generalize the contraction method, originally proposed by Elgot and Rabin and later extended by Carton and Thomas, from labeled linear orderings to colored deterministic trees. The method we propose rests on a suitable notion of indistinguishability of trees with respect to tree automata that allows us to reduce a number of instances of the acceptance problem for tree automata to decidable instances involving regular trees. We prove that such a method works effectively for a large class of trees, which is closed under noticeable operations and includes all the deterministic trees of the Caucal hierarchy obtained via unfoldings and inverse finite mappings as well as several trees outside such a hierarchy.
Keywords
deterministic automata; finite automata; trees (mathematics); Caucal hierarchy; acceptance problem; colored deterministic trees; contraction method; decide MSO theories; inverse finite mappings; labeled linear orderings; monadic second-order logic; tree automata; Automata; Binary trees; Cost accounting; Logic; Periodic structures; Specification languages; Tree data structures; Vegetation mapping;
fLanguage
English
Publisher
ieee
Conference_Titel
Logic in Computer Science, 2007. LICS 2007. 22nd Annual IEEE Symposium on
Conference_Location
Wroclaw
ISSN
1043-6871
Print_ISBN
0-7695-2908-9
Type
conf
DOI
10.1109/LICS.2007.6
Filename
4276559
Link To Document