THE PRE-MAPPING SENTENCE PLANNER IN MULTILINGUAL NATURAL LANGUAGE GENERATION SYSTEM

Natural language generation is the research area that studies how the computer can be used to generate natural language text. Its classic architecture is a pipeline structure consisting of three parts: macroplanner, microplanner and surface generator. Analyzed in this paper is the microplanner in the ACNLG, a multilingual nature language generation system, and the concept “sentence optimizer based on language resource pre mapping” is put forward. It does not only optimize the sentences, but also maps the language independent content into each language’s resource, and the processing should follow the instruction of the language resource. This sentence optimizer has all the ability of the classic method, and solves the problem that in the multilingual environment it is hard to handle the nuance of each language. It can optimize the sentences according to the feature of the target language, and increase the ability of the sentence optimizer greatly.