電視新聞內容分類與索引之研究

Abstract

隨著資訊科技的快速發展，資訊的儲存、呈現、處理與交換的方式 亦產生很大的影響，最根本的改變就是文件與影音的電子化；再加 上網路的普及，使得各種檢索系統的使用情況也更為頻繁。新一代 資訊檢索系統，尤其是多媒體查詢的系統，必須能夠運用更具效率 的自動化技術，以提供簡易有效的檢索服務。 本論文主要是使用影像處理、光學字元辨識等技術，搭配我們所提 出的文件分類、關鍵字擷取及視訊與文件關聯等技術，將各段電視 新聞影片與相關之文字新聞自動建立聯結，接著再對各則新聞作分 類與建立關鍵字索引，以此建立一個多媒體新聞索引。

With the rapid development of information technologies, the way people store, present, process and exchange data have been changed. More and more documents, including videos and audio, are store in digital formats now. The population of internet also increases the frequency of using searching systems to find information we want. The new types of information retrieval systems, especially multi-media query systems, must use more efficient automatic techniques to provide easy and effective retrieval service. In this thesis, a method is proposed which utilize current text information processing techniques for news video classification. The goal is to build a system which effectively analysis news videos and further classify them into some previously defined categories. The method proposed use recent probabilistic based text classifying algorithms associate with optical character recognition techniques. Experimental results show that this method can achieve good classification performance under real conditions using daily TV news.