假新聞依據型態，動機或寫作風格有著不同的寫法。先前的假新聞欺騙檢測的相關研究使用人工方式提取特徵，這樣的方法受限於人類自身的語言理解能力。即便這樣，假新聞中內的語言變異特徵的提取還是有其困難。在本研究中，我們探討了使用自動提取重要語義特徵方法的可能性。這些被提取的語意特徵不受限於人類本身的語言理解，同時我們也探討是否這些方法可以捕獲演變中的語言變異性。我們的實驗結果顯示，我們的模型可以與使用傳統機器學習並由人工進行特徵篩選的模型達到相當的效果。

Fake news articles are differently written, depending on the type, motivation and writing style. Previous work in deception detection in fake news use features that are manually made and are limited to predefined human understandings of linguistics. That being said, it is difficult to extract the shifts in linguistic variability in fake news articles. In this work, we investigate the possibility of using a method that will be able to automatically extract important linguistic-based features. The extracted linguistic-features are not limited to our understandings of linguistics, and we will investigate if they can to capture evolving linguistic variability in fake news. Our experimental results show that our model achieves results that are comparable to the models that use traditional machine learning, which are limited to manual feature selection.

1 Introduction ...... 1
2 Related work ...... 5
2.0.1 Knowledge-based Fake News Analysis . . . . . . . . . . . . . 5
2.0.2 Propagation-based Fake News Analysis . . . . . . . . . . . . . 6
2.0.3 Style-based Fake News Analysis . . . . . . . . . . . . . . . . . 7
3 Proposed Method 9
3.0.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.0.2 Data Pre-processing . . . . . . . . . . . . . . . . . . . . . . . 11
3.0.3 Graph Construction . . . . . . . . . . . . . . . . . . . . . . . 13
3.0.4 Linguistic Patterns Extraction . . . . . . . . . . . . . . . . . . 15
3.0.5 Model Training . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4 Experimental Results .......... 25
5 Conclusions ........ 33
References .......35

[1] Fake news: What exactly is it–and how can you spot it? (2019). https://www.telegraph.co.uk/technology/0/fake-news-exactly-has- really-had-influence/.
[2] Collins 2017 word of the year shortlist (2017).
https://www.collinsdictionary.com/word-lovers-blog/new/collins -2017-word-of-the-year-shortlist,396,HCB.html.
[3] Read all about it: The biggest fake news stories of 2016 (2016).
https://www.cnbc.com/2016/12/30/read-all-about-it-the-biggest-fa ke-news-stories-of-2016.html.
[4] fake news’threat to media; editorial decisions, outside actors at fault (2018). https://www.monmouth.edu/polling-institute/reports/monmouthpoll_ us_040218/.
[5] James Mahon. The definition of lying and deception. Stanford Encyclopedia of Philosophy, 12 2015.
[6] G. Feng, V. Hirst. Detecting deceptive opinion with profile compatibility. 03 2013.
[7] David M Markowitz and Jeffrey T Hancock. Linguistic traces of a scientific fraud: The case of diederik stapel. PloS one, 9(8):e105937, 2014.
[8] Yla R Tausczik and James W Pennebaker. The psychological meaning of words: Liwc and computerized text analysis methods. Journal of language and social psychology, 29(1):24–54, 2010.
[9] Arthur C Graesser, Danielle S McNamara, Max M Louwerse, and Zhiqiang Cai. Coh-metrix: Analysis of text on cohesion and language. Behavior research methods, instruments, & computers, 36(2):193–202, 2004.
[10] Xinyi Zhou and Reza Zafarani. Fake news: A survey of research, detection methods, and opportunities. CoRR, abs/1812.00315, 2018.
[11] A platform for civil, fact-based, and engaging discussions. https://fiskkit.com.
[12] Sarah Cohen, James T. Hamilton, and Fred Turner. Computational journalism. Commun. ACM, 54:66–71, 10 2011.
[13] Ndapandula Nakashole and Tom M. Mitchell. Language-aware truth assessment of fact candidates. volume 1, pages 1009–1019, 06 2014.
[14] Giovanni Luca Ciampaglia, Prashant Shiralkar, Luis Rocha, Johan Bollen, Filippo Menczer, and Alessandro Flammini. Computational fact checking from knowledge networks. PloS one, 10, 01 2015.
[15] Hannah Bast, Björn Buchhold, and Elmar Haussmann. Relevance scores for triples from type-like relations. In Proceedings of the 38th International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR ’15, pages 243–252, New York, NY, USA, 2015. ACM.
[16] Zilong Zhao, Jichang Zhao, Yukie Sano, Orr Levy, Hideki Takayasu, Misako Takayasu, Li Daqing, and Shlomo Havlin. Fake news propagate differently from real news even at early stages of spreading. 03 2018.
[17] Federico Monti, Fabrizio Frasca, Davide Eynard, Damon Mannion, and Michael Bronstein. Fake news detection on social media using geometric deep learning. 02 2019.
[18] Dina Pisarevskaya. Rhetorical structure theory as a feature for deception detection in news reports in the russian language. 06 2017.
[19] Svitlana Volkova, Kyle Shaffer, Jin Yea Jang, and Nathan Hodas. Separating facts from fiction: Linguistic models to classify suspicious and trusted news posts on twitter. pages 647–653, 01 2017.
[20] Gary D. Bond, Rebecka D. Holman, Jamie-Ann Eggert, Lassiter Speller, Olivia N. Garcia, Sasha C. Mejia, Kohlby W. Mcinnes, Eleny C. Ceniceros, and Rebecca Rustige. ‘lyinapos; ted’, ‘crooked hillary’, and ‘deceptive donald’: Language of lies in the 2016 us presidential debates. Applied Cognitive Psychology, 31, 10 2017.
[21] Martin Potthast, Johannes Kiesel, Kevin Reinartz, Janek Bevendorff, and Benno Stein. A stylometric inquiry into hyperpartisan and fake news. CoRR, abs/1702.05638, 2017.
[22] Elvis Saravia, Hsien-Chi Toby Liu, and Yi-Shin Chen. Deepemo: Learning and enriching pattern-based emotion representations. 04 2018.
[23] Sepp Hochreiter and Jürgen Schmidhuber. Long short-term memory. Neural computation, 9:1735–80, 12 1997.