Results 1  10
of
248
Training Tree Transducers
 IN HLTNAACL
, 2004
"... Many probabilistic models for natural language are now written in terms of hierarchical tree structure. Treebased modeling still lacks many of the standard tools taken for granted in (finitestate) stringbased modeling. The theory of tree transducer automata provides a possible framework to ..."
Abstract

Cited by 128 (11 self)
 Add to MetaCart
(Show Context)
Many probabilistic models for natural language are now written in terms of hierarchical tree structure. Treebased modeling still lacks many of the standard tools taken for granted in (finitestate) stringbased modeling. The theory of tree transducer automata provides a possible framework to draw on, as it has been worked out in an extensive literature. We motivate the use of tree transducers for natural language and address the training problem for probabilistic treetotree and treetostring transducers.
Statistical syntaxdirected translation with extended domain of locality
 In Proc. AMTA 2006
, 2006
"... A syntaxdirected translator first parses the sourcelanguage input into a parsetree, and then recursively converts the tree into a string in the targetlanguage. We model this conversion by an extended treetostring transducer that have multilevel trees on the sourceside, which gives our system m ..."
Abstract

Cited by 119 (14 self)
 Add to MetaCart
(Show Context)
A syntaxdirected translator first parses the sourcelanguage input into a parsetree, and then recursively converts the tree into a string in the targetlanguage. We model this conversion by an extended treetostring transducer that have multilevel trees on the sourceside, which gives our system more expressive power and flexibility. We also define a direct probability model and use a lineartime dynamic programming algorithm to search for the best derivation. The model is then extended to the general loglinear framework in order to rescore with other features like ngram language models. We devise a simpleyeteffective algorithm to generate nonduplicate kbest translations for ngram rescoring. Initial experimental results on EnglishtoChinese translation are presented. 1
Regular Tree and Regular Hedge Languages over Unranked Alphabets: Version 1
, 2001
"... We survey the basic results on regular tree languages over unranked alphabets; that is, we use an unranked alphabet for the labels of nodes, we allow unbounded, yet regular, degree nodes and we treat sequences of trees that, following Courcelle, we call hedges. The survey was begun by the first ..."
Abstract

Cited by 113 (5 self)
 Add to MetaCart
We survey the basic results on regular tree languages over unranked alphabets; that is, we use an unranked alphabet for the labels of nodes, we allow unbounded, yet regular, degree nodes and we treat sequences of trees that, following Courcelle, we call hedges. The survey was begun by the first and third authors. Subsequently, when they discovered that the second author had already written a summary of this view of tree automata and languages, the three authors decided to join forces and produce a consistent review of the area. The survey is still unfinished because we have been unable to find the time to finish it. We are making it available in this unfinished form as a research report because it has, already, been heavily cited in the literature.
Symbolic model checking with rich assertional languages
 Theoretical Computer Science
, 1997
"... Abstract. The paper shows that, by an appropriate choice of a rich assertional language, it is possible to extend the utility of symbolic model checking beyond the realm of bddrepresented nitestate systems into the domain of in nitestate systems, leading to a powerful technique for uniform veri c ..."
Abstract

Cited by 113 (4 self)
 Add to MetaCart
Abstract. The paper shows that, by an appropriate choice of a rich assertional language, it is possible to extend the utility of symbolic model checking beyond the realm of bddrepresented nitestate systems into the domain of in nitestate systems, leading to a powerful technique for uniform veri cation of unbounded (parameterized) process networks. The main contributions of the paper are a formulation of a general framework for symbolic model checking of in nitestate systems, a demonstration that many individual examples of uniformly veri ed parameterized designs that appear in the literature are special cases of our general approach, verifying the correctness of the Futurebus+ design for all singlebus con gurations, extending the technique to tree architectures, and establishing that the presented method is a precise dual to the topdown invariant generation method used in deductive veri cation. 1
An Overview of Probabilistic Tree Transducers for Natural Language Processing
, 2005
"... Probabilistic finitestate string transducers (FSTs) are extremely popular in natural language processing, due to powerful generic methods for applying, composing, and learning them. Unfortunately, FSTs are not a good fit for much of the current work on probabilistic modeling for machine translati ..."
Abstract

Cited by 71 (6 self)
 Add to MetaCart
Probabilistic finitestate string transducers (FSTs) are extremely popular in natural language processing, due to powerful generic methods for applying, composing, and learning them. Unfortunately, FSTs are not a good fit for much of the current work on probabilistic modeling for machine translation, summarization, paraphrasing, and language modeling. These methods operate directly on trees, rather than strings. We show that tree acceptors and tree transducers subsume most of this work, and we discuss algorithms for realizing the same benefits found in probabilistic string transduction.
Catching Bugs in the Web of Program Invariants
, 1996
"... MrSpidey is a userfriendly, interactive static debugger for Scheme. A static debugger supplements the standard debugger by analyzing the program and pinpointing those program operations that may cause runtime errors such as dereferencing the null pointer or applying nonfunctions. The program anal ..."
Abstract

Cited by 70 (26 self)
 Add to MetaCart
MrSpidey is a userfriendly, interactive static debugger for Scheme. A static debugger supplements the standard debugger by analyzing the program and pinpointing those program operations that may cause runtime errors such as dereferencing the null pointer or applying nonfunctions. The program analysis of MrSpidey computes value set descriptions for each term in the program and constructs a value flow graph connecting the set descriptions. Using the set descriptions, MrSpidey can identify and highlight potentially erroneous program operations, whose cause the programmer can then explore by selectively exposing portions of the value flow graph.
Monadic secondorder definable graph transductions: A survey
 TCS
, 1994
"... Courcelle, B., Monadic secondorder definable graph transductions: a survey, Theoretical Computer Science 126 (1994) 5375. Formulas of monadic secondorder logic can be used to specify graph transductions, i.e., multivalued functions from graphs to graphs. We obtain in this way classes of graph tr ..."
Abstract

Cited by 59 (8 self)
 Add to MetaCart
(Show Context)
Courcelle, B., Monadic secondorder definable graph transductions: a survey, Theoretical Computer Science 126 (1994) 5375. Formulas of monadic secondorder logic can be used to specify graph transductions, i.e., multivalued functions from graphs to graphs. We obtain in this way classes of graph transductions, called monadic secondorder definable graph transductions (or, more simply, d&able transductions) that are closed under composition and preserve the two known classes of contextfree sets of graphs, namely the class of hyperedge replacement (HR) and the class of vertex replacement (VR) sets. These two classes can be characterized in terms of definable transductions and recognizable sets of finite trees, independently of the rewriting mechanisms used to define the HR and VR grammars. When restricted to words, the definable transductions are strictly more powerful than the rational transductions such that the image of every finite word is finite; they do not preserve contextfree languages. We also describe the sets of discrete (edgeless) labelled graphs that are the images of HR and VR sets under definable transductions: this gives a version of Parikh’s theorem (i.e., the characterization of the commutative images of contextfree languages) which extends the classical
Macro Tree Transducers, Attribute Grammars, and MSO Definable Tree Translations
 Inform. and Comput
, 1998
"... A characterization is given of the class of tree translations definable in monadic second order logic (MSO), in terms of macro tree transducers. The first main result is that the MSO definable tree translations are exactly those tree translations realized by macro tree transducers (MTTs) with reg ..."
Abstract

Cited by 56 (18 self)
 Add to MetaCart
(Show Context)
A characterization is given of the class of tree translations definable in monadic second order logic (MSO), in terms of macro tree transducers. The first main result is that the MSO definable tree translations are exactly those tree translations realized by macro tree transducers (MTTs) with regular lookahead that are single use restricted. For this the single use restriction known from attribute grammars is generalized to MTTs. Since MTTs are closed under regular lookahead, this implies that every MSO definable tree translation can be realized by an MTT. The second main result is that the class of MSO definable tree translations can also be obtained by restricting MTTs with regular lookahead to be finite copying, i.e., to require that each input subtree is processed only a bounded number of times. The single use restriction is a rather strong, static restriction on the rules of an MTT, whereas the finite copying restriction is a more liberal, dynamic restriction on the ...
Decidable Approximations of Sets of Descendants and Sets of Normal Forms
, 1997
"... We present here decidable approximations of sets of descendants and sets of normal forms of Term Rewriting Systems, based on specific tree automata techniques. In the context of rewriting logic, a Term Rewriting System is a program, and a normal form is a result of the program. Thus, approximations ..."
Abstract

Cited by 50 (14 self)
 Add to MetaCart
(Show Context)
We present here decidable approximations of sets of descendants and sets of normal forms of Term Rewriting Systems, based on specific tree automata techniques. In the context of rewriting logic, a Term Rewriting System is a program, and a normal form is a result of the program. Thus, approximations of sets of descendants and sets of normal forms provide tools for analysing a few properties of programs: we show how to compute a superset of results, to prove the sufficient completeness property, or to find a criterion for proving termination under a specific strategy, the sequential reduction strategy.