## Compositions of Top-down Tree Transducers with "-rules

### Cached

### Download Links

Citations: | 2 - 2 self |

### BibTeX

@MISC{Maletti_compositionsof,

author = {Andreas Maletti and Heiko Vogler},

title = {Compositions of Top-down Tree Transducers with "-rules},

year = {}

}

### OpenURL

### Abstract

Abstract. Top-down tree transducers with "-rules ("tdtt) are a restricted version of extended top-down tree transducers. They are implemented in the framework Tiburon and ful ll some criteria desirable in a machine translation model. However, they compute a class of transformations that is not closed under composition (not even for linear and nondeleting "tdtt). A composition construction that composes "tdtt M and N is presented. It is correct whenever (i) M has at most one output symbol in each rule, (ii) M is deterministic or N is linear, and (iii) M is total or N is nondeleting. This corresponds nicely to a classical composition result by Baker. 1

### Citations

3836 |
Introduction to automata theory, languages and computation. Addison-Wesley publishing company
- John, Ullman
- 1979
(Show Context)
Citation Context ...nicely to a classical composition result by Baker. 1 Introduction Many aspects of machine translation (MT) of natural languages can be formalized by employing weighted nite-state (string) transducers =-=[1, 2]-=-. Successful implementations based on this phrase-based approach are, e.g., the At&t Fsm toolkit [3], Xerox's nite-state calculus [4], Rwth toolkit [5], Carmel [6], and OpenFst [7]. However, this phra... |

633 | Tree adjoining grammars
- Joshi
- 2003
(Show Context)
Citation Context ...f such formalisms are, e.g., the top-down tree transducer [8, 9], the extended top-down tree transducer [10{12], the synchronous tree substitution grammar [13], the synchronous tree-adjoining grammar =-=[14]-=-, the multi bottom-up tree transducer [15{18], and the extended multi bottom-up tree transducer [19]. Some of these models are formally compared in [20, 21, 19] and an overview on their usability in s... |

334 | Regular models of phonological rule systems
- Kaplan, Kay
- 1994
(Show Context)
Citation Context ...ormalized by employing weighted nite-state (string) transducers [1, 2]. Successful implementations based on this phrase-based approach are, e.g., the At&t Fsm toolkit [3], Xerox's nite-state calculus =-=[4]-=-, Rwth toolkit [5], Carmel [6], and OpenFst [7]. However, this phrase-based approach is not expressive enough, e.g., to handle the rotation needed in the translation of the English sentence structure ... |

235 |
Tree Automata. Akadémiai Kiadó
- Gécseg, Steinby
- 1984
(Show Context)
Citation Context ... that a formalism for MT should have are identi ed in [25, 26]. These are expressivity, modularity, inclusiveness, and trainability. This list was later extended to include preservation of regularity =-=[27, 28]-=-. None of the above mentioned formalisms ful lls all 5 criteria. In this paper, we consider top-down tree transducers with "-rules ("tdtt). Such transducers are a restricted version of extended top-do... |

235 | Transductions and Context-Free languages - Berstel - 1979 |

120 |
Mappings and grammars on trees
- Rounds
- 1970
(Show Context)
Citation Context ... by storing the subject, which might be very long, in its finite memory. Syntax-based (or tree-based) formalisms can remedy this shortage. Examples of such formalisms are the top-down tree transducer =-=[8, 9]-=-, the extended topdown tree transducer [10–12], the synchronous tree substitution grammar [13], the synchronous tree-adjoining grammar [14], the multi bottom-up tree transducer [15–18], and the extend... |

103 | Training Tree Transducers
- Graehl, Knight, et al.
(Show Context)
Citation Context ...p-down tree transducers and as such implemented in Tiburon. They can express rotation (using copying), are inclusive (in the sense that they can simulate nite-state string transducers), are trainable =-=[12]-=-, and preserve regularity (if we disallow copying). We prove a Baker-like composition result [29, 30] saying that the composition of two "tdtt M and N can be computed by one "tdtt if (i) M has at most... |

98 | The Design Principles of a Weighted Finite-State Transducer Library
- Mohri, Pereira, et al.
- 2000
(Show Context)
Citation Context ...guages can be formalized by employing weighted finite-state (string) transducers [1, 2]. Successful implementations based on this word- or phrase-based approach are, for example, the At&t Fsm toolkit =-=[3]-=-, Xerox’s finite-state calculus [4], the Rwth toolkit [5], Carmel [6], and OpenFst [7]. However, the phrase-based approach is not expressive enough, for example, to easily handle the rotation needed i... |

71 | OpenFst: A general and efficient weighted finite-state transducer library
- Allauzen, Riley, et al.
- 2007
(Show Context)
Citation Context ...]. Successful implementations based on this word- or phrase-based approach are, for example, the At&t Fsm toolkit [3], Xerox’s finite-state calculus [4], the Rwth toolkit [5], Carmel [6], and OpenFst =-=[7]-=-. However, the phrase-based approach is not expressive enough, for example, to easily handle the rotation needed in the translation of the English structure NP-V-NP (subject-verb-noun phrase) to the A... |

64 |
An overview of probabilistic tree transducers for natural language processing
- Knight, Graehl
- 2005
(Show Context)
Citation Context ...sducer [15{18], and the extended multi bottom-up tree transducer [19]. Some of these models are formally compared in [20, 21, 19] and an overview on their usability in syntax-based MT is presented in =-=[22, 23]-=-. For example, the system Tiburon [24] implements weighted extended top-down tree transducers. ? This author was nancially supported by the Ministerio de Educacion y Ciencia (MEC) grant JDCI-2007-760.... |

56 |
Generalized2 sequential machine maps
- Thatcher
(Show Context)
Citation Context ...the pronoun, which might be very long, in its nite memory. Formalisms that are based on syntax (or trees) can remedy this shortage. Examples of such formalisms are, e.g., the top-down tree transducer =-=[8, 9]-=-, the extended top-down tree transducer [10{12], the synchronous tree substitution grammar [13], the synchronous tree-adjoining grammar [14], the multi bottom-up tree transducer [15{18], and the exten... |

48 | A decoder for syntax-based statistical MT
- Yamada, Knight
- 2002
(Show Context)
Citation Context ...ry of the model; extended top-down tree transducers have full symmetry in the linear and nondeleting case. It is often beneficial in the development process to train “small” task-specific transducers =-=[26]-=-. Then we would like to compose those “small” transducers to obtain a single transducer, to which further operations can be applied. The success of the finite-state transducer toolkits (like Carmel [6... |

35 |
Composition of top-down and bottom-up tree transductions
- Baker
- 1979
(Show Context)
Citation Context ...ing), are inclusive (in the sense that they can simulate nite-state string transducers), are trainable [12], and preserve regularity (if we disallow copying). We prove a Baker-like composition result =-=[29, 30]-=- saying that the composition of two "tdtt M and N can be computed by one "tdtt if (i) M has at most one output symbol in each rule, (ii) M is deterministic or N is linear, and (iii) M is total or N is... |

31 | 2006b. Tiburon: A weighted tree automata toolkit
- May, Knight
(Show Context)
Citation Context ...tom-up tree transducer [19]. Some of these models are formally compared in [20, 21, 19] and an overview on their usability in syntax-based MT is presented in [22, 23]. For example, the system Tiburon =-=[24]-=- implements weighted extended top-down tree transducers. ? This author was nancially supported by the Ministerio de Educacion y Ciencia (MEC) grant JDCI-2007-760.Four properties that a formalism for ... |

30 | The power of extended top-down tree transducers
- Maletti, Graehl, et al.
- 2009
(Show Context)
Citation Context ...2007-760 and MEC grant MTM-2007-63422.In this paper, we consider top-down tree transducers with ε-rules (εtdtts), which are a (syntactically) restricted version of extended top-down tree transducers =-=[25, 21]-=- and as such implemented in Tiburon [24]. In fact, εtdtts properly generalize finite-state transducers, and thus existing models for the phrase-based approach can be straightforwardly translated into ... |

27 | Capturing practical natural language transformations
- Knight
(Show Context)
Citation Context ...p-down tree transducers. ? This author was nancially supported by the Ministerio de Educacion y Ciencia (MEC) grant JDCI-2007-760.Four properties that a formalism for MT should have are identi ed in =-=[25, 26]-=-. These are expressivity, modularity, inclusiveness, and trainability. This list was later extended to include preservation of regularity [27, 28]. None of the above mentioned formalisms ful lls all 5... |

26 |
Bottom-up and top-down tree transformations—A comparison
- Engelfriet
- 1975
(Show Context)
Citation Context ...ing), are inclusive (in the sense that they can simulate nite-state string transducers), are trainable [12], and preserve regularity (if we disallow copying). We prove a Baker-like composition result =-=[29, 30]-=- saying that the composition of two "tdtt M and N can be computed by one "tdtt if (i) M has at most one output symbol in each rule, (ii) M is deterministic or N is linear, and (iii) M is total or N is... |

24 | FSA: an efficient and flexible C++ toolkit for finite state automata using on-demand computation
- Kanthak, Ney
- 2004
(Show Context)
Citation Context ...te (string) transducers [1, 2]. Successful implementations based on this word- or phrase-based approach are, for example, the At&t Fsm toolkit [3], Xerox’s finite-state calculus [4], the Rwth toolkit =-=[5]-=-, Carmel [6], and OpenFst [7]. However, the phrase-based approach is not expressive enough, for example, to easily handle the rotation needed in the translation of the English structure NP-V-NP (subje... |

22 | Synchronous grammars as tree transducers
- Shieber
- 2004
(Show Context)
Citation Context ...ar [13], the synchronous tree-adjoining grammar [14], the multi bottom-up tree transducer [15{18], and the extended multi bottom-up tree transducer [19]. Some of these models are formally compared in =-=[20, 21, 19]-=- and an overview on their usability in syntax-based MT is presented in [22, 23]. For example, the system Tiburon [24] implements weighted extended top-down tree transducers. ? This author was nanciall... |

20 |
M.: Tree Languages, in: Handbook of Formal Languages
- Gécseg, Steinby
- 1997
(Show Context)
Citation Context ... that a formalism for MT should have are identi ed in [25, 26]. These are expressivity, modularity, inclusiveness, and trainability. This list was later extended to include preservation of regularity =-=[27, 28]-=-. None of the above mentioned formalisms ful lls all 5 criteria. In this paper, we consider top-down tree transducers with "-rules ("tdtt). Such transducers are a restricted version of extended top-do... |

18 | Bitransduction de forêts - Arnold, Dauchet - 1976 |

17 | Compositions of extended top-down tree transducers
- Maletti
- 2008
(Show Context)
Citation Context ...ar [13], the synchronous tree-adjoining grammar [14], the multi bottom-up tree transducer [15{18], and the extended multi bottom-up tree transducer [19]. Some of these models are formally compared in =-=[20, 21, 19]-=- and an overview on their usability in syntax-based MT is presented in [22, 23]. For example, the system Tiburon [24] implements weighted extended top-down tree transducers. ? This author was nanciall... |

12 | Propriétés de clôture d’une extension de transducteurs d’arbres déterministes - Lilin - 1981 |

11 | Une Généralisation des Transducteurs D’états Finis D’arbres: les S-transducteurs. Thèse 3ème cycle, Université de Lille - Lilin - 1978 |

10 | M.: Transductions inversibles de forêts. Thèse 3ème cycle M. Dauchet, Université de Lille - Arnold, Dauchet - 1975 |

10 | H.: A bottom-up characterization of deterministic top-down tree transducers with regular look-ahead - Fülöp, Kühnemann, et al. - 2004 |

9 | H.: Linear deterministic multi bottom-up tree transducers - Fülöp, Kühnemann, et al. - 2005 |

9 | Extended multi bottom-up tree transducers: Composition and decomposition
- Engelfriet, Lilin, et al.
- 2009
(Show Context)
Citation Context ...cer [10{12], the synchronous tree substitution grammar [13], the synchronous tree-adjoining grammar [14], the multi bottom-up tree transducer [15{18], and the extended multi bottom-up tree transducer =-=[19]-=-. Some of these models are formally compared in [20, 21, 19] and an overview on their usability in syntax-based MT is presented in [22, 23]. For example, the system Tiburon [24] implements weighted ex... |

9 |
Full abstract families of tree series I
- Kuich
- 1999
(Show Context)
Citation Context ...this section by showing that ln-εTOP is not closed under composition. This shows that linear, nondeleting εtdtts are strictly less expressive than linear, nondeleting recognizable tree transducers of =-=[32]-=- because the latter compute a class of transformations that is closed under composition [32, Theorem 2.4]. Note that the latter have regular extended right-hand sides. Theorem 5. ln-εTOP ; ln-εTOP ̸⊆ ... |

7 |
Weighted automata algorithms, in: Handbook of Weighted Automata
- Mohri
(Show Context)
Citation Context ...nicely to a classical composition result by Baker. 1 Introduction Many aspects of machine translation (MT) of natural languages can be formalized by employing weighted nite-state (string) transducers =-=[1, 2]-=-. Successful implementations based on this phrase-based approach are, e.g., the At&t Fsm toolkit [3], Xerox's nite-state calculus [4], Rwth toolkit [5], Carmel [6], and OpenFst [7]. However, this phra... |

3 |
M.: OpenFst: A general and e cient weighted nite-state transducer library
- Allauzen, Riley, et al.
- 2007
(Show Context)
Citation Context ...ing) transducers [1, 2]. Successful implementations based on this phrase-based approach are, e.g., the At&t Fsm toolkit [3], Xerox's nite-state calculus [4], Rwth toolkit [5], Carmel [6], and OpenFst =-=[7]-=-. However, this phrase-based approach is not expressive enough, e.g., to handle the rotation needed in the translation of the English sentence structure PRO-V-NP (pronoun-verb-noun phrase) to the Arab... |

3 |
Mathematical and Computational Aspects of Lexicalized Grammars
- Shabes
- 1990
(Show Context)
Citation Context ...r trees) can remedy this shortage. Examples of such formalisms are, e.g., the top-down tree transducer [8, 9], the extended top-down tree transducer [10{12], the synchronous tree substitution grammar =-=[13]-=-, the synchronous tree-adjoining grammar [14], the multi bottom-up tree transducer [15{18], and the extended multi bottom-up tree transducer [19]. Some of these models are formally compared in [20, 21... |

2 |
The design principles of a weighted nite-state transducer library
- Mohri, Pereira, et al.
- 2000
(Show Context)
Citation Context ...MT) of natural languages can be formalized by employing weighted nite-state (string) transducers [1, 2]. Successful implementations based on this phrase-based approach are, e.g., the At&t Fsm toolkit =-=[3]-=-, Xerox's nite-state calculus [4], Rwth toolkit [5], Carmel [6], and OpenFst [7]. However, this phrase-based approach is not expressive enough, e.g., to handle the rotation needed in the translation o... |

2 |
Applications of Weighted Automata
- Knight, May
- 2009
(Show Context)
Citation Context ...sducer [15{18], and the extended multi bottom-up tree transducer [19]. Some of these models are formally compared in [20, 21, 19] and an overview on their usability in syntax-based MT is presented in =-=[22, 23]-=-. For example, the system Tiburon [24] implements weighted extended top-down tree transducers. ? This author was nancially supported by the Ministerio de Educacion y Ciencia (MEC) grant JDCI-2007-760.... |

1 |
H.: Fsa: an e cient and exible C ++ toolkit for nite state automata using on-demand computation
- Kanthak, Ney
(Show Context)
Citation Context ...ying weighted nite-state (string) transducers [1, 2]. Successful implementations based on this phrase-based approach are, e.g., the At&t Fsm toolkit [3], Xerox's nite-state calculus [4], Rwth toolkit =-=[5]-=-, Carmel [6], and OpenFst [7]. However, this phrase-based approach is not expressive enough, e.g., to handle the rotation needed in the translation of the English sentence structure PRO-V-NP (pronoun-... |

1 |
finite-state toolkit. ISI/USC
- Graehl
- 1997
(Show Context)
Citation Context ...d nite-state (string) transducers [1, 2]. Successful implementations based on this phrase-based approach are, e.g., the At&t Fsm toolkit [3], Xerox's nite-state calculus [4], Rwth toolkit [5], Carmel =-=[6]-=-, and OpenFst [7]. However, this phrase-based approach is not expressive enough, e.g., to handle the rotation needed in the translation of the English sentence structure PRO-V-NP (pronoun-verb-noun ph... |

1 |
Requirements on an MT system. personal communication
- Knight
- 2008
(Show Context)
Citation Context ...p-down tree transducers. ? This author was nancially supported by the Ministerio de Educacion y Ciencia (MEC) grant JDCI-2007-760.Four properties that a formalism for MT should have are identi ed in =-=[25, 26]-=-. These are expressivity, modularity, inclusiveness, and trainability. This list was later extended to include preservation of regularity [27, 28]. None of the above mentioned formalisms ful lls all 5... |

1 |
Composition of linear and nondeleting top-down tree transducers with ε-rules
- Heger
- 2008
(Show Context)
Citation Context ...ly seen that, for the linear and nondeleting "tdtt of Example 8, there exists an equivalent linear and nondeleting "tdtt that is maximally outputseparated (see Example 10). It was formally veri ed in =-=[32]-=- that for every linear and nondeleting "tdtt there exists an equivalent linear and nondeleting "tdtt that is maximally output-separated. In general, every "tdtt can be transformed into an equivalent m... |

1 |
Compositions of tree transducers. personal communication
- Knight
- 2008
(Show Context)
Citation Context ...ion is immediate by Theorems 12 and 18. Strictness follows from the fact that transformations of "TOP need not preserve recognizability whereas those of ln "TOP do [27, 31]. ut Open problem: The task =-=[33]-=- is to identify suitably complex classes C of tree transformations and P1; : : : ; Pn; Q1; : : : ; Qn C such that for every 1; 2 2 C we have _ ( 1 2 Pi) ^ ( 2 2 Qi) () 1 ; 2 2 C : 1 i n In fact, the c... |

1 |
Requirements on an MT system
- Knight
- 2008
(Show Context)
Citation Context ...lizes Baker’s composition result for nondeterministic top-down tree transducers. Moreover, our investigation on composition of εtdtt might give some insight in how to solve the open problem stated in =-=[34, 35]-=-: find a tree transducer model that is expressive, modular, inclusive, and trainable. Another open problem (stated by one of the referees) is the following: given a tree transducer type (such as linea... |