網路發展日新月異，不僅多數新聞媒體網站已經發展成熟，網路新聞資料庫、訂閱服務或是對話機器人等新興的網路新聞服務也不斷推出，人們接收新聞的速度透過這些服務也大幅提升，在提升搜尋新聞的方便性的同時，也容易因大量的相似資訊而造成資訊過載。
為了解決上述的問題，我們透過抽取並分析新聞內的主詞-動詞-受詞事件，並且基於廣義知網以及命名實體辨識，針對作為主詞或受詞的名詞片語進行分類，建立新聞事件結構。接著，我們透過新聞分類分群建立上層樹狀結構，在底層，我們透過時間、地點及事件結構連結不同新聞內的事件串成新聞事件鏈，形成樹狀新聞知識圖譜。
最後，我們基於上述樹狀新聞知識圖譜建立一個新聞聊天對話系統，提供使用者結構化的新聞資訊。

By the high development of internet, almost all the news media websites are mature. Many internet news services are released, such as internet news database, web news subscription or chat bot. The speed of people accepting information and news is more fast than before via these services. Despite high convenience of searching data, it might cause information overload.
To solve the problem, we extract the subject-verb-object events of news. We classify the subjects and the object of events via E-Hownet and Name Entity Recognition (NER). We use class of subject and object to build event structure. Then, we make use of news classification and clustering to build tree-structure layer. Under the tree-structure layer, we build event chain by time, location and event structure. We build tree-structure knowledge graph after that.
Finally, we build a news dialogue system based on the tree-structure knowledge graph. It provides structured news information.

[1] “2019 台灣網路報告,” 2019. [線上]. Available: https://report.twnic.tw/2019/assets/download/TWNIC_TaiwanInternetReport_2019_CH.pdf.
[2] “自由時報,” 自由時報, [線上]. Available: https://www.ltn.com.tw/.
[3] “蘋果日報首頁,” 蘋果日報, [線上]. Available: https://tw.appledaily.com/home/.
[4] “Google trends,” [線上]. Available: https://trends.google.com.tw/trends/explore?date=2010-01-01%202020-08-27&q=chatbot.
[5] Delia Rusu, Jozef Stefan Institute and Jozef Stefan International Postgraduate School, James Hodson, Anthony Kimball, Unsupervised Techniques for Extracting and Clustering Complex Events, 2014.
[6] Chen Lin, Chun Lin, Jingxuan Li, Dingding Wang, Yang Chen, Tao Li, “Generating event storylines from microblogs,” 於 ACM international conference, 2012.
[7] Katsiaryna Stalpouskaya, Christian Baden, “To Do or Not to Do: the Role of Agendas for Action in Analyzing News Coverage of Violent Conflict,” 於 ACL, 2015.
[8] 許孟淵 黃純敏, 以本體論為基礎之新聞事件檢索與瀏覽, 2006.
[9] Y.-H. Cheng, News Bot: A Conversational News Service Based On Event-centric Knowledge Graphs and E-HowNet, 2017.
[10] Jacob Devlin, Ming-Wei Chang, Kenton Lee, Kristina Toutanova, BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding, 2019.
[11] Chuan-Jie Lin, Hsin-Hsi Chen, Che-Chia Liu, Jin-He Tsai, Hong-Jia Wong, Open-Domain Question Answering on Heterogeneous Data, 2001.
[12] David M. Blei, Andrew Y. Ng, Michael I. Jordan, “Latent Dirichlet Allocation,” Journal of Machine Learning Research 3, 2003.
[13] Gevorg Poghosyan and Georgiana Ifrim, Real-time News Story Detection and Tracking with Hashtags, 2016.
[14] “E-HowNet,” [線上]. Available: http://ehownet.iis.sinica.edu.tw/.
[15] “聯合新聞網,” 聯合報, [線上]. Available: https://udn.com/news/index.
[16] “CkipTagger,” [線上]. Available: https://github.com/ckiplab/ckiptagger.
[17] Daniel Ramage, David Hall, Ramesh Nallapati and Christopher D. Manning, Labeled LDA: A supervised topic model for credit attribution in multi-labeled corpora, Computer Science Department, Stanford University.
[18] Remi Bois ´ and Guillaume Gravier, Eric Jamet and Maxime Robert, Emmanuel Morin, Pascale Sebillot, “Language-based Construction of Explorable News Graphs for Journalists,” 於 20th International Conference on Computational. Linguistics pages 50–56, 2010.
[19] Rishab Goel, Seyed Mehran Kazemi, Marcus Brubaker, Pascal Poupart, Diachronic Embedding for Temporal Knowledge, 2019.
[20] Zhaohui Wu, Chen Liang, C. Lee Giles, The Pennsylvania State University, Storybase: Towards Building a Knowledge Base for News Events, 2015.