MACHINE TRANSLATION The world is growing smaller every day. Accurate communication between countries, people, and businesses is becoming more and more important. Manual machine translation has become too slow and cumbersome, and computers have taken over this essential but difficult task. Here are some of the challenges inherent in automated translation, and ways that today's sophisticated hardware and software are dealing with, and resolving, the many problems.
Diagram: Parsing and Interlingua Representation: Simple parsing (a) yields a single parse tree even if the sentence is ambiguous (i.e., it can be parsed several ways). At this stage, the parsing is purely syntactic. Sophisticated parsing (b) yields a parse forest composed of all parses that the grammar allows. Syntactically, the sentence in this example can be interpreted several ways. Semantic analysis of the parse forest will yield a most likely interpretation (syntactic reading #2), which becomes the interlingua representation. An interlingua representation (c) details the syntax of a sentence and includes enough semantics to increase the likelihood of creating an accurate synthesis. Elements of the representation are actually coded as numbers that are indexes to multilingual dictionary entries and phrase structure templates.
Photograph: The Intergraph machine-language Translator system provides a GUI, although the underlying technology is based on an earlier command-oriented system.
Peter M. Benton, formerly McGraw-Hill's chief scientist, evaluated and applied advanced technologies, including automated translation and other natural language processing systems. He now consults on the assessment and introduction of new technologies. He can be reached on BIX as ``benton.''
Per comentaris i observacions sobre el servidor, poseu-vos en contacte amb l'Administrador WWW