基于领域本体的新闻视频检索

新闻视频作为视频数据中有代表性的一种媒体,受到人们的广泛关注,对新闻视频的检索要求也越来越高.传统的新闻视频检索大多是非语义层面的,采用的是基于关键词的检索方法，难于获得令人满意的查准率和查全率．本文提出一种基于领域本体的新闻视频检索框架，定义了新闻视频检索中的新闻视频对象，使用语义表达能力强的领域本体来指导视频语义对象的标注，并针对“一词多义”问题提出了“概念域-概念”两阶段概念消歧算法；针对自然语言检索问题，使用领域本体进行查询优化和查询扩展，并提出了查询语句自动生成方法．实验表明，基于领域本体的新闻视频检索方法可以有效的提高检索性能．

News Video Retrieval Using Domain Ontology

News video is one of the most representative multimedia information, which has being received more and more attention. Users need a great improvement of the performance of news video retrieval. Traditionally, news video retrieval is based on keyword matching technologies without considering the semantics. So it is hard to get satisfying recall and precision. Architecture of news video retrieval based on domain ontology is proposed in this paper. News video object (NVO for short) is defined, and the annotation of NVO is directed with domain ontology. "Region-Concept" Disambiguation algorithm is suggested to solve the problem of ambiguity. A technique of query expansion and optimization is performed using domain ontology, and a method of automatic generation for SQL query is put forward. Based on the above technologies, ontology-based retrieval is realized in some degree in news video. The evaluations show that the effectiveness of news video retrieval can be improved with the ontology-based retrieval method.