【作者】 刘东波；

【导师】 高春鸣；

【作者基本信息】 湖南师范大学 ， 计算机软件与理论， 2009， 硕士

【摘要】 当前基于文本的检索技术和基于内容的检索技术广泛应用于多媒体资源检索领域。然而,基于文本的检索技术采用手工标注,自动化程度低,标注质量因人而异,不但费时费力,而且难以保证标注结果语义的一致性;基于内容的检索技术虽然可在一定程度上提高检索的自动化程度,然而其内容描述只是针对多媒体数据的底层特征,没有表达资源高层语义信息。另一方面,传统的信息检索模型仅从字面意义上进行关键词匹配,缺乏语义处理能力,存在信息的误检、漏检等缺陷。基于本体的信息检索以及潜在语义索引方法是实现语义检索的两种途径,能够提高检索的查全率和查准率。本文首先阐述本体相关理论,分析现有的信息检索技术和模型,以及基于本体的信息检索实现机制;然后提出基于本体的潜在语义索引模型,给出本体与潜在语义索引结合的具体实现过程,并构建基于本体的查询扩展算法实现对用户查询的语义扩展,进一步提高语义检索的准确性和有效性;最后通过实地考察动画制作流程,分析并设计基于本体的动画素材检索系统,提出面向动画素材的语义检索系统框架和实现途径,完成系统分析和总体设计,以及本体服务器、查询请求预处理器、查询请求重构器的设计和实现,并通过分析实验结果说明语义检索能够有效的提高检索性能。更多还原

【Abstract】 Currently, the text-based retrieval technology and content-based retrieval technology are widely used in the field of retrieval of multimedia resources. However, the text-based search technology uses manual annotation which leads to the low degree of automation. The quality of annotating varies from person to person, this technology is not only time-consuming and laborious but also difficult to ensure semantic consistency of the result of annotating. Although the content-based retrieval technology can improve automation degree of retrieval at a certain extent, the description of contents only directs to the physical characteristics of multimedia data, and doesn’t descirbe the high-level semantic information of resources. On the other hand, traditional information retrieval models only match from the literal meaning of words, short of the capacity to deal with semantic information, and having defects such as false information retrieval and omission of undetected information. Ontology-based information retrieval and latent semantic indexing (LSI) method are two ways to realize semantic retrieval and can improve retrieval recall ratio and precision ratio.Firstly, this paper set out the ontological theory, and analyze implementation mechanism of Ontology-based information retrieval as well as the existed information retrieval techniques and models. Secondly, we propose a latent semantic indexing model based on ontology, giving the process of the combination between ontology and LSI in detail, building query expansion algorithm based on ontology to realize the semantic expansion of user query which further improves the accuracy and effectiveness of semantic retrieval. Finally, through the survey of animation production flow, we analyze and design an ontology-based retrieval system for animation materials, proposing the framework and implementation method of the semantic retrieval system, completing design and implementation of the ontology server , request pre-processor and request re-constructor in addition to system analysis and system design, and indicating through the analysis of experimental results that semantic retrieval can effectively improve retrieval performance.更多还原