網際網路新聞文章心情偵測之研究

Abstract

全世界每天有數以萬計的新聞被報導，在這些新聞裡，僅有少部份與自己有 關的，大多數是毫不相關的。隨著網際網路使用者數量大幅度增加，網路已取代 傳統媒體成為最受矚目的大眾媒體，如何從網路上眾多的新聞之中，以最短時間 去篩選出自己需要的、喜愛的及完全不需閱讀的新聞乃是一個值得關注的議題。

本研究會以預測讀者閱讀新聞後的心情為目標，使用Yahoo!奇摩新聞的心 情投票資料，透過CKIP 的斷詞切字處理，計算出每個詞彙的Log Likelihood Ratio 值，與其心情比例分數結合之後排序篩選，找出優秀的特徵值作為分類依 據，最後再放入LibSVM 分類建構出模型，預測讀者閱讀新聞後可能呈現的心情 狀況，並進一步設計出關鍵詞彙挑選系統，供讀者在選擇閱讀新聞時參考。

There's millions and thousands news coming out everyday. Only limited number of these news are relevant to a particular person. In the digital era, Internet has surpassed traditional media and become one of the most attractive media. How do we effective and efficiently filter through the huge amount of information on the Internet for finding those pieces of information which we need, like or we don't need to read? There's millions and thousands news coming out everyday. Only limited number of these news are relevant to a particular person. In the digital era, Internet has surpassed traditional media and become one of the most attractive media. How do we effective and efficiently filter through the huge amount of information on the Internet for finding those pieces of information which we need, like or we don't need to read?