资源描述:
《气象落区文本自动生成研究》由会员上传分享,免费在线阅读,更多相关内容在应用文档-天天文库。
1、ComputerEngineeringandApplications计算机工程与应用2014,50(13)247气象落区文本自动生成研究12323324吴焕萍,吕终亮,张华平,罗兵,高健,李笑侃,何国豪,王永超1232332WUHuanping,LVZhongliang,ZHANGHuaping,LUOBing,GAOJian,LIXiaokang,HEGuohao,4WANGYongchao1.国家气候中心,北京1000812.国家气象中心,北京1000813.北京理工大学,北京1000814.中国地质大学,北京1000831.Nation
2、alClimateCenter,Beijing100081,China2.NationalMeteorologicalCenter,Beijing10081,China3.BeijingInstituteofTechnology,Beijing100081,China4.ChinaUniversitiesofGeosciences,Beijing100083,ChinaWUHuanping,LVZhongliang,ZHANGHuaping,etal.Textgenerationonweatherfallingareadescription.
3、Com-puterEngineeringandApplications,2014,50(13):247-251.Abstract:Thetextcontentneedsthehighqualitycharacteristicofprecision,efficiency,logicalnessandaswellasnaturallanguageexpressioninordertogeneratetextdescriptionautomaticallyforweatherforecastandservicefields,howevercurre
4、ntlyitlacksmorematuretechniquewaystosolveallproblems.Basedontheanalysisofthecontentof“weatherpublicreport”documents,itproposesaframeworkandits’processingoftextgenerationonweatherfallingareadescriptionbyintroducingofGeographicInformationScience(GIS)andNaturalLanguageProcessi
5、ng(NLP)science.Feathermore,itfocusesonthefourkeyissuesforfurtherdiscussion,suchasinformationextractionandconceptualmodelbuildingfrommassivehistorydocuments,geospatialareapartitioning,meteorologicalelementsspatial-temporalreasoning,textorganizationandgenerationandpost-proces
6、sing.Meanwhile,itproposesthetechniqueimplementationindetailfortheseissues.Com-paredwiththecontentbetweenforecastermanuscriptandcomputertext,theresultsshowthegeneraltextgenerationmethodhasapotentialapplicationprospectingivenmeteorologyforecastandservicesfield.Keywords:Natura
7、lLanguageProcessing(NLP);textfeatureextraction;meteorologicaldataspatialanalysis;textauto-generation摘要:面向天气预报和气象服务的文本内容的计算机自动或者半自动生成方法,对文本生成质量要求较高,即要准确、高效、合理,还需要符合自然语言表达,存在较多技术问题。在深入分析中央气象台每日发布的“天气公报”文本内容的基础上,结合地理信息科学和自然语言处理科学方法提出了面向气象落区文本语言生成的基本原理与流程,重点从历史文本内容分析与特征提取、地理区域划
8、分、气象要素空间分析、文本组织与生成等关键技术问题进行了深入讨论,并给出了相应的技术实现。计算机自动生成结果与预报员人工撰写的文本内容对比分析也较好地证明了面向特地