客製化影音檢索系統之研發—以可移式語音機之設計為例

Abstract

面對資訊爆炸的時代，傳統的關鍵字比對搜尋技術將無法更有效率地提供使用者之所需，而通訊與資訊等相關產業，在各數位系統匯流成單一網路的狀況下，如何主動提供用戶所需的數位內容，也就成為研究重要的課題。 本研究提出一個創新的平台架構，可以從網路上獲得免費的影音內容，且同時解決頻寬、收視習慣、合法性、及自動下載等問題。主要特點在於此平台不是直接傳送內容檔案給用戶端，而是只傳送一個標準協議的檔案推薦超鏈結位置及相關註記資訊給用戶端即可，隨後它會自動去瀏覽並自動下載該目標影音檔，且按照Script的描述編成各頻道節目及播放順序，如此，讓網路上各式各樣的節目，瞬間成為各個產業可合法應用的內容。 本平台將藉由一個自動推薦系統，將合適的影音檔案推薦給用戶，其中使用模糊的屬性權重資訊檢索技術，來改善傳統推薦系統使用明確比對的方式，用戶對影音偏好的屬性，僅需利用語言變數來描述表達其重要性，便可輕鬆獲得理想的影音檔案推薦。為了驗證提出之服務模式，本研究設計出一個互動式的數位可移式語音機，使理論與實務相結合，讓使用者能藉此系統得到良好的有聲內容推薦。

Face the times of information explosion, traditional keyword search technique will not provide an efficiently approach to the users’ need. Besides, the related industries such as communication and information technology etc., will under the situation which digital information remit to single network flow. How provide the digital content to match the customers’ needs, will also become the research important topic. This research presents a creative platform structure, can acquire free information and legal of video content from the Internet. In advance, it also can resolve the bandwidth bottleneck, watching habit, and automatic download problem etc. Main distinguishing feature's lying in this platform isn't a direct transmission content the file carry for customer, but the file which only delivers a standard agreement recommends a super chain knot position and related note to record information to the customer carry. Moreover, it will browse automatically and download the target video file automatically later, and apply Script description to become each channel program and broadcast by proper order. Therefore, the various programs will be became legal content from the Internet and used by the industry fields. This platform will construct an auto recommend system, and recommend the suitable video file to the customers. Besides, it uses the fuzzy attributes weight for the information as an index technique among them. The goal is to improve the traditional recommendation system to use an explicit approach. In our approach, the users only describe the attribute of video hobby and expression of variable importance, can acquire an ideal video file recommendation easily. This research designs an interaction digital learning machine, making theories and actual situation combine together, let the users be able to get to have a content to recommend goodly by this system.