## The reactive engine for modular transducers (2005)

Venue: | Algebra, Meaning and Computation, Essays Dedicated to Joseph A. Goguen on the Occasion of His 65th Birthday |

Citations: | 10 - 9 self |

### BibTeX

@INPROCEEDINGS{Huet05thereactive,

author = {Gérard Huet and Benoît Razet},

title = {The reactive engine for modular transducers},

booktitle = {Algebra, Meaning and Computation, Essays Dedicated to Joseph A. Goguen on the Occasion of His 65th Birthday},

year = {2005},

pages = {355--374}

}

### OpenURL

### Abstract

Abstract. This paper explains the design of the second release of the Zen toolkit [5–7]. It presents a notion of reactive engine which simulates finite-state machines represented as shared aums [8]. We show that it yields a modular interpreter for finite state machines described as local transducers. For instance, in the manner of Berry and Sethi, we define a compiler of regular expressions into a scheduler for the reactive engine, chaining through aums labeled with phases — associated with the letters of the regular expression. This gives a modular composition scheme for general finite-state machines. Many variations of this basic idea may be put to use according to circonstances. The simplest one is when aums are reduced to dictionaries, i.e. to (minimalized) acyclic deterministic automata recognizing finite languages. Then one may proceed to adding supplementary structure to the aum algebra, namely non-determinism, loops, and transduction. Such additional choice points require fitting some additional control to the reactive engine. Further parameters are required for some functionalities. For instance, the local word access stack is handy as an argument to the output routine in the case of transducers. Internal virtual addresses demand the full local state access stack for their interpretation. A characteristic example is provided, it gives a complete analyser for compound substantives. It is an abstraction from a modular version of the Sanskrit segmenter presented in [9]. This improved segmenter uses a regular relation condition relating the phases of morphology generation, and enforcing the correct geometry of morphemes. Thus we obtain compound nouns from iic*.(noun+iic.ifc), where iic and ifc are the respectively prefix and suffix substantival forms for compound formation. 1 Regular morphology Dedicated to Joseph Goguen for his 65th birthday We first consider the simplest framework for finite automata, where the state transition graph is a dictionary structure (lexical tree or trie). Such structures represent acyclic deterministic finite-state automata, with maximal sharing of initial paths. Every state is accessible from the initial state, and we may also assume that every state is on an accepting path. When we minimize the tree as a dag, we obtain the corresponding minimal deterministic automaton. Such

### Citations

142 | Morphology and Computation - Sproat - 1992 |

102 |
From regular expressions to deterministic automata
- Berry, Sethi
- 1986
(Show Context)
Citation Context ...e, which chains the morphemes dictionary lookup with transitions corresponding to the regular expression recognizer. We shall use for this setup variants of the compiling algorithm of Berry and Sethi =-=[2]-=-. 1.1 Automaton interface We use as algorithmic description language Pidgin ML, a core applicative subset of Objective Caml. Thus our algorithms may be read as rigorous higher-order inductive definiti... |

64 | Partial derivatives of regular expressions and finite automaton constructions
- Antimirov
- 1996
(Show Context)
Citation Context ...rrectness criteria are invariant with the permutation of choices induced by priority selection according to these weights. 4.4 A variant using Antimirov’s compiling algorithm V. Antimirov proposed in =-=[1]-=- another algorithm for compiling regular expressions, using a notion of partial derivative. This algorithm produces automata that may be significantly smaller than the ones obtained by the Berry-Sethi... |

60 | Finite-State Language Processing - Roche, Shabes - 1997 |

25 |
A functional toolkit for morphological and phonological processing, application to a Sanskrit tagger
- Huet
- 2005
(Show Context)
Citation Context ... their interpretation. A characteristic example is provided, it gives a complete analyser for compound substantives. It is an abstraction from a modular version of the Sanskrit segmenter presented in =-=[9]-=-. This improved segmenter uses a regular relation condition relating the phases of morphology generation, and enforcing the correct geometry of morphemes. Thus we obtain compound nouns from iic*.(noun... |

17 | Local languages and the Berry–Sethi algorithm, Theoret - Berstel, Pin - 1996 |

13 | contexts and the sharing functor: Techniques for symbolic computation - Linear - 2003 |

10 | The Zen computational linguistics toolkit
- Huet
- 2002
(Show Context)
Citation Context ... algorithms may be read as rigorous higher-order inductive definitions, while being directly executable, in the spirit of literate programming. We first recall the basic structures of the Zen toolkit =-=[5]-=-. We use as basic alphabet the natural numbers provided by the hardware processor: module Word : sig type l e t t e r = i n t and word = l i s t l e t t e r ; end ; Thus the basic morphology operation... |

10 | The Zen computational linguistics toolkit: Lexicon structures and morphology computations using a modular functional programming language - Huet - 2002 |

5 | Automata Mista
- HUET
- 2003
(Show Context)
Citation Context ...rance Abstract. This paper explains the design of the second release of the Zen toolkit [5–7]. It presents a notion of reactive engine which simulates finite-state machines represented as shared aums =-=[8]-=-. We show that it yields a modular interpreter for finite state machines described as local transducers. For instance, in the manner of Berry and Sethi, we define a compiler of regular expressions int... |