多语种自然语言生成系统中的语言模型

介绍了在多语种自然语言生成系统中如何用统一的模型来表示各种语种的语言知识。本文将语言知识分为语义资源和语法资源两个部分。前者通过Ｓｃｈｅｍａ和优化规则确定文本的内容；后者根据子结构类，句法规则和词典确定文本的具体形式。该模型以复杂特征集为语言知识的载体，将表示抽象语义的数据扩展为生成不同语种文本所需的语言资源。

Language Model for Multilingual Natural Language Generation

This paper introduced a knowledge representation model for multilingual natural language text generation system. It can be divided into two levels: semantic resources and syntax resources. The former is used in describing text content by schema and optimized rules; the latter is used in sentence pattern constructing, language resource mapping and determining text specific form according to the sentence structure class, syntactic rules and lexical information. The model is based on a complex feature set. It can be used to extend the abstractinitial semantic data to all kinds of language resource for multilingual text generation.