tailieunhanh - Báo cáo khoa học: "Syntax and Interpretation"

This paper describes a model of syntactical analysis. It shows first how a context-free grammar formalism can be modified in order to reduce the number of rules required by a natural language, and second, how this formalism can be extended to the handling of some non-contextfree phenomena. The description of the algorithm is also given. | Mechanical Translation and Computational Linguistics June 1966 Syntax and Interpretation by B. Vauquois G. Veillon and J. Veyrunes . Grenoble France This paper describes a model of syntactical analysis. It shows first how a context-free grammar formalism can be modified in order to reduce the number of rules required by a natural language and second how this formalism can be extended to the handling of some non-context-free phenomena. The description of the algorithm is also given. The process of discontinuous constituents is performed by transformations. Introduction In a mechanical translation system based on a succession of models where each model represents a level of the source language or of the target language one must establish their linkage. In the phase of analysis related to the source language the linkage paths are called directed upward as the successive models correspond to hierarchical levels which are more and more elevated in the language whereas in the phase of synthesis related to the target language the linkage paths are directed downward. The analysis of a text of the source language L consists in finding within the model of the highest level called model M3 a formula or a sequence of formulas the representation of which in L is the given text. If each formula of M3 is represented in L by at least one sentence of L we have a so-called sen-tence-by-sentence analysis. All sentences of L which are the different representations of one and the same formula of M3 are called equivalent with respect to the model. Every sentence of L which is a representation that two or more formulas of M3 have in common is called ambiguous in the model. The model M3 will be admitted to have also a representation in the chosen target language L but we do not bother to know whether M3 possesses still further representations in the languages L L etc. Under these conditions each non-ambiguous sentence of L can be made to correspond to a sentence of L .