預測新聞報導對讀者情緒的影響

計畫名稱:預測新聞報導對讀者情緒的影響

所屬單位:資訊系

研究團隊:自然語言處理實驗室

計畫主持人:陳信希

研究人員:林欣毅

資源需求:Perl, gcc, g++, LibSVM, SVM-Light

使用期間:2008/03~

研究主題:
預測新聞報導對讀者情緒的影響

研究內容概述:
We study the emotion classification of news articles from readers’ perspective. The method of using support vector machines (SVM) on different combinations of character bigram, word, news metadata (e.g., news category, news agency, hour of publication, news reporter, and news event location) and word emotion features is explored. The experimental materials consist of Chinese news articles downloaded from Taiwan Yahoo! News, which allows a reader to vote for one of eight emotions of happiness, sadness, anger, surprising, boredom, heartwarming, novelty, and usefulness. Because a news article affects different readers differently, each article contains a distribution of emotions. For classification purpose, the most popular emotion in an article is chosen as the correct emotion class.

詳細計畫內容 回到上一頁