【作者】 吴鹏；

【导师】 黄少滨；

【作者基本信息】 哈尔滨工程大学 ， 计算机应用技术， 2011， 硕士

【摘要】 复杂网络的研究已经成为互联网、社会学、生物学等多个领域的一个基础课题。节点的聚集现象是很多复杂网络具有的一个特性,被称为簇、社团或群。语义复杂网络的聚集现象可以揭示网络中节点间语义的相似性,据此可以从少量节点的精确语义获得其他节点的潜在语义。首先,语义复杂网络是一种加权的复杂网络,以往的研究主要是为发现非加权复杂网络的聚集现象,本文将研究一种加权复杂网络即语义复杂网络的聚集算法从而发掘节点间的语义关系。这种加权复杂网络是将用户检索的反馈转化为描述对象语义关系的复杂网络。为此提出一种基于语义核的凝聚型层次聚类算法CACNSC,发现不同粒度下语义的聚集现象。然后,通过推荐少量对象由专家标注,或根据少量对象的已有的精确语义,再基于层次聚类过程中构建的语义关系树,实现其他对象的语义标注。对含有大量噪音的模拟反馈数据和Princeton Shape Benchmark的真实反馈信息的实验表明,所提方法在语义聚集和标注两个方面都取得了较好的效果。最终,将本文提出的算法应用的实际的三维模型检索系统中,随着用户反馈信息的增加所构建的语义复杂网络节点间的语义会更加丰富,不但提高CACNSC算法结果的准确率,这样也可以使得语义更加准确,提高相关检索效率。更多还原

【Abstract】 Complex networks has become the Internet, sociology, biology and other fields of a basic task. Aggregation node is a feature in many of complex networks, known as clusters or groups. The aggregation of complex networks can reveal many potential problems and also reflect the relationship between network nodes. Previous studies found that non-weighted aggregation of complex networks. This thesis will examine the complex network that is the weighted complex network aggregation semantics.First, this thesis studies the weighted complex network cluster. This complex networks converts various feedbacks into the semantic complex network. The thesis proposes an agglomerative hierarchical clustering method based on semantic core, named CACNSC, to analyze the semantic accumulation under different granularities.Second, the thesis shows a mechanism to recommend a few important multimedia objects for authority annotation and states the automatically annotation method using semantic information of limited multimedia objects. The proposed method is verified by the ideal feedbacks with high noise and the real feedbacks of Princeton Shape Benchmark. The method performs quite well not only in semantic clustering but also in annotation.Finally, this thesis implements a three-dimensional model retrieval system. With the increase in user feedback, the semantics of nodes will be more abundant. This can make more precise semantics to improve the retrieval efficiency.更多还原