多语种自然语言生成系统中的预映射句子规划器

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

自然语言生成是研究如何用计算机来生成自然语言文本的研究领域 ,经典结构是宏观规划、微观规划和表层生成的流水线结构 .分析了多语种自然语言生成系统 ACNL G中的微观规划器 ,提出了“基于语言资源预映射的句子优化器”,它不仅对句子进行优化 ,而且负责将语种无关的文章内容映射到具体语种的表达方式中 .其核心思想就是将处理过程同具体语种的语言资源相分离 ,但同时又以语言资源为导向 ,既继承了现有方法的优点 ,同时也解决了多语种条件下各语种之间细微差别的处理问题 ,使句子优化能够针对目标语种的特点进行优化操作 ,从而使优化效果有实质的提高 。

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.