TED關聯圖的建構：最佳化方法之運用

Abstract

TED.com是全球知名視頻網站。本研究以最佳化方法設計TED關聯圖和視覺檢索框架。此研究首先計算TED中video-video，topic-topic，video- topic 和topic-speaker等資訊間的距離和相似度；再用最佳化技術計算這些TED資訊投影在二維空間中的位置座標；然後使用視覺化方法呈現資訊間的關聯。TED關聯圖展示了TED資訊間的關聯關係與分群聚類，為使用者提供一全新的資訊組織、搜索與結果展示的模式。

TED.com is a highly influential online video website. This study proposes an optimized method of designing TED relation maps and building an interactive retrieval framework. The proposed method first computes the distance and similarity between TED different objects such as video-video, topic-topic, video- topic and topic-speaker. Then, calculate the coordinates of these multi-faceted and multi-level TED objects that mapping in two-dimensional space by optimized method. It allows users get and understand TED visual metaphors and knowledge quickly by browse and filter. TED relation maps display the relationships about TED videos, topics and speakers, which provides users a new mode about information organization, search and results display.