【作者】 周开朋；

【导师】 韩敏；

【作者基本信息】 大连理工大学 ， 控制理论与控制工程， 2006， 硕士

【摘要】 湿地是自然界最富生物多样性的生态景观和人类社会赖以生存和发展的环境之一,在提供水资源、均化洪水、调节气候、保护生物多样性等方面发挥了重要的作用,因而被誉为“自然之肾”。目前人们对湿地保护的认识十分欠缺,而且不同的组织和机构对湿地信息及湿地知识的不同理解或不同的表达格式,严重地影响了相互之间的信息共享和互操作。本文将本体技术应用到湿地保护中,首先分析了影响湿地资源的各因素及其产生的生态特征变化,创建湿地保护输入输出需求驱动模型,其次介绍了该模型的各知识集,并结合“骨架法”建立湿地保护领域本体,最终确定了湿地领域中的概念及概念之间的关系,一定程度上实现湿地信息共享和重用,从而逐步达到人们对湿地保护的共识。 本文通过将湿地保护输入输出需求驱动模型及湿地保护领域本体融入到湿地信息源(文档集)中,并结合对湿地文档集的概念性、关系性、预测性特点及文档标引源位置特性的分析,研究建立了概念性子文档、关系性子文档及改进的文档向量空间模型。以此用来计算文档的概念权值和关系权值,进而量化湿地文档所含的知识信息。而针对文档的预测性采用神经网络方法实现,在现有的文档特征基础上预测文档的级别,实现文档集文档之间的排序。 采用C/S架构设计后,依据文档集分级评测方法,使用Java+Oracle开发技术,进行信息源评测系统的设计与实现,完成了从实现方法到系统实现的跨越。系统主要包括初集本体设定、模拟文档产生、概念提取及关系分析等8个模块,并给出了各模块的实现原理和步骤。在系统中,通过提取文档集的词频、文档标引源位置特性及本体关系距离等初集本体特征,作为样本集,并采用BP神经网络预测出文档集内文档的分级排序。更多还原

【Abstract】 The wetland is the richest landscape in zoology variety and one of the most important environment resources for human. It plays an important role in supplying the water resource, meaning the flood, modulating the climate and protecting the variety of life forms. So the wetland is honored as "the kidney of nature". At present, most people are short of wetland protection consciousness, and it is different organization’s different understanding and expressing format that badly affects information share and interoperation in wetland. According to systemic research and analyses on each factor of wetland protection through ontology technology, the need drive model on input and output of wetland protection is created, the knowledge set of it is introduced, then the wetland domain ontology is established combining Skeletal Methodology. So, the concepts and relations of wetland can be established definitely, the share and recycle of wetland information can be realized at a certain extent, and then wetland protection recognition of people is consentaneous by step.The conceptive document, relative document and the improved Vector Space Model (VSM) are established, through putting the input and output need drive model and the wetland protection domain ontology into wetland information resources (document set), and combining with the analyses of marked sources location and wetland document set characters as conceptive, relative, and predictive. Then the concept and relation weights are computed, the knowledge and information of wetland document can be quantified. The prediction of document is realized by Neural Network (NN), then the document level can be predicted, and the documents of the set can be graded based on the document existing character.In this paper, after the design of C/S Integrated Framework, Java and Oracle technology are used to establish the document set evaluating system combining with the document set evaluating methods. The system is composed of eight modules, such as initial ontology setting, simulating document production, concept distilling and relation analyzing, and so on. Each module is designed and realized in detail in this paper. Word frequency, marked sources location and ontology relation distances handled by the system are used as input samples of the Back-Propagation Neural Networks (BPNN) to predict the document level.更多还原