冬奥气象服务文本自动生成模型研究

AUTOMATIC TEXT GENERATION MODEL OF OLYMPIC WINTER GAMES METEOROLOGICAL SERVICE

目前的冬奥气象服务文本依赖手工撰写,存在生成效率不高、内容提取不够精准和英文自然语言表达不足的问题。在深入分析冬奥服务文本特征基础上,提出以自然语言生成理论为指导,研究冬奥数据和知识构建特征引擎,为文本生成提供内容规划;结合功能合一语法(Functional unification grammar,FUG),开展递归运算,实现从短语到句式生成;利用XML Schema对文本篇章、段落结构进行组织管理。经模型自动生成的冬奥气象服务文本在2019年至2020年现场服务团队的冬训工作得以应用,为其他大型冬季赛事气象服务保障提供了技术方案。

At present,the meteorological service text of the Olympic Winter Games relies on manual writing,with the problems of low generation efficiency,inaccurate content extraction and insufficient natural language expression in English.Based on the in-depth analysis of the text features of the Olympics Winter Games service,it was proposed to study the Olympics Winter Games data and knowledge to build a feature engine under the guidance of the natural language generation theory,so as to provide content planning for text generation.Combined with the FUG,the recursive operation was carried out to realize the generation from phrases to sentences.XML schema was used to organize and manage the text and paragraph structure.The Olympic Winter Games meteorological service text automatically generated by the model has been applied in the winter training of the on-site service team from 2019 to 2020,providing a technical scheme for the meteorological service guarantee of other large-scale winter events.