請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/5915
標題: | 公共議程新聞的自動化議題探勘 Mining Salient Issues from News Articles on Public Agendas |
作者: | Chia-Mau Ni 倪嘉懋 |
指導教授: | 許永真(Jane Yung-jen Hsu) |
關鍵字: | 新聞,議題探勘,公共議程,新聞自動摘要,子議題探勘,信息抽取, news,issue mining,public agenda,news summarization,subtopic mining,information extraction, |
出版年 : | 2013 |
學位: | 碩士 |
摘要: | 公共議程(Public Agenda)係指一系列相關議題,引發公共討論並引起社會注意。在公共議程中,受關注的議題對於議程的發展方向有較高的影響力。然而,議題的顯著程度可能受到媒體的刻意操作,偏頗的媒體往往將違反其立場或利益的議題邊緣化。
本論文介紹一個透過分析新聞文章中的引述句以在公共議程中自動化探勘議題的方法。新聞中的引述句記錄了公眾人物與領域專家的意見,我們藉由將引述句依主題分群,以辨識出新聞文章中被大量辯論的議題。為了增進引述句的分群效果,我們提出「議題顯著詞」來代表每個引述句的主題。 我們收集了一份關於「核四停建」公共議程的新聞文章語料庫以驗證方法的效能,該公共議程是 2013 年台灣社會高度關注的議程之一。我們人工標記出新聞文章中的顯著議題,並將引述句依議題分類。我們使用這份語料庫來驗證引述句分群與議題探勘的效能。 A public agenda is a set of issues or concerns that merit public attention. The issues that attract a lot of public attention are influential to the direction of the public agenda. However, the salience of issues might be purposely transferred by the media. Biased news organizations marginalize or filter issues that are against their positions or private interests. In this thesis we propose a method to automatically mine salient issues from news articles. Quotations in news articles describe comments from public figure and domain experts. Our method for issue mining is based on quotation analysis. By clustering quotations according to their subjects, we identify issues that are widely debated on. We introduce issue significant terms to improve the performance of the method. To evaluate the performance of issue mining, we compile a corpus of news articles about the public agenda on Lungmen Nuclear Power Plant. The public agenda is a focus for concern in Taiwan in 2013. We manually identify the ground truth of salient issues in the corpus and categorize quotations according to these issues. The performance of quotation clustering and issue mining is evaluated with the ground truth. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/5915 |
全文授權: | 同意授權(全球公開) |
顯示於系所單位: | 資訊工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-102-1.pdf | 857.21 kB | Adobe PDF | 檢視/開啟 |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。