發展社群媒體事件之圖像快篩方法：以2016年美濃地震之Twitter資料為例 - 政大學術集成

全文

(1)Department of Computer Science National Chengchi University. Master’s Thesis. 3 2016. 立. 4 治政大 Twitter. ‧ 國. 學. ‧. Developing a Fast Screening Method for Visual Images in Social Media Events: A Case Study of Twitter Data during 2016 Meinong Earthquake n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. July 2018. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(2) 4 Twitter. 3. Developing a Fast Screening Method for Visual Images in Social Media Events: A Case Study of Twitter Data during 2016 Meinong Earthquake Student: FENG, SHU-CHAO Advisor: Kung Chen. 立. 政治大. er. io. sit. y. ‧. ‧ 國. 學. Nat. A Thesis. n. submitted toaDepartment of Computer i v Science. l C n U h e n g c University i National Chengchi h. in partial fulfillment of the Requirements for the degree of Master in Computer Science. July 2018 DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(3) 4 3 2016. Twitter. 3. t. 0 /3. ,d. 0. 5. .. -. t a. 0. 3 3. r3. ‧. ‧ 國. r3. 學. r3. y. Nat. io. 3. al. n. meta data. 4. .. sit. 3 2016. t. 政治 9大. 立. d. r3. er. 3. Ch. engchi. i n U. v. 3. 3. 3. 8 5 r 9. 2016. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(4) Developing a Fast Screening Method for Visual Images in Social Media Events: A Case Study of Twitter Data during 2016 Meinong Earthquake Abstract Recently, thanks to various social media platforms and availability of mobile web, people have got used to interactive through the internet anywhere anytime. Activities on social media have become everyone's routine, such as searching or sharing information and communication. In the meanwhile, these activities make social media platforms a treasure for getting information. This boundless, real-time information strongly connected to our. 政治大 able to analyze specific events of the real world. This feature is especially important at 立 analyzing disasters. At the beginning of disasters, getting information is vital for those. real life, which means, by locating the information of some position and moment, we are. ‧ 國. 學. authorized. By properly extract messages from social media, experts can realize more and much quickly about the disaster to take the next steps.. ‧. This work provides a case study of 2016 Meinong Earthquake happened in Taiwan by. Nat. sit. y. exploring the data on Twitter. First of all, this work analyzes the metadata of tweets and. er. io. show that tweets with images are more likely to be retweeted. Then, after using the computer vision services to label the images, this work provides the results and resistance. n. al. Ch. i n U. v. of using label co-occurrence to cluster images. In the end, by crawling the user information. engchi. of popular tweets' publisher, we can realize that besides the media of Taiwan, there are also media from other countries caring about this disasters. Moreover, we also know that most personal user publisher of popular tweets use Japanese. The reason might be that Japanese users are the second most in Tweet and location of Japan is relatively near Taiwan than other occidental countries.. keywords : 2016 Meinong Earthquake, Twitter, Social media analysis, Text analysis, Graph theory. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(5) t. r. t. 3 3. 3. y. PLSM Lab 6. 立. 0. 政 bug治大 4. ‧. 3. -. r. io. sit. y. Nat. 3. t. 學. ‧ 國. 6. t9. 5. n. al. er. 3. 3. 00. Ch. engchi a. i n U. v. t. 0. M. -. o. 10. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(6) t. 立. v. ‧. ‧ 國. 學. y. 政治大. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. 1 1 2 4 6 6 7 8 8 9 10 11 11 12 13 15 15 17 18 22 23 26 29 37 37 39 39 42 44 51 60. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(7) l 0 t. 0 3 t. t. 0. t. t. 0. 立. 3. 學. ‧ 國. 4. 政治大. 7. 0. y. sit. n. al. er. io. t. Nat. t. ‧. 0. Ch. engchi. i n U. v. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(8) Diffbot. Twitter. How-Old.net. 3. 立5 5. TZ3 2016-02-15 02. 5. Nat. TZ4 2016-02-19 21. sit. y. 5 u. n. al. er. io. 3. ‧. TZ2 2016-02-07 18. 學. ‧ 國. TZ1 2016-02-06 06. 政治大. d. Ch. engchi. i n U. v. d AntiDupl.NET. UI. Google. 30. MS AWS IBM Clarifi. 30 30 30 30. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(9) t t. 0. t. 2. 立. 政治大. ‧. ‧ 國. 學 y. sit. n. al. er. .. io. 12. Nat. i. 25 2 3. Ch. engchi. i n U. v. 0. 10 4 9 28 r. 4. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(10) r. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(11) t. 7 t. 8. 7 y. w. 8. 立 7. 7. ‧. ‧ 國. 學. [1]. 政治大. n. al. er. io. sit. y. Nat w. 8. Ch. e n g c”h i. i n U. v. ~. [2]. 8. 7 t. — 7. 7 y. ~. 8 y. 8. 1 Daily time spent on social networking by internet users worldwide from 2012 to 2017 in (minutes) https://www.statista.com/statistics/433871/daily-social-media-usage-worldwide/ 1. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(12) “A picture is worth a thousand words.”. t. Twitter. 8 Diffbot page classifier ~v. 2012. Twitter w Twitter. w Twitter. t. Twitter. Tweet. t. 140. 8 治政大. 立. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. 1-1 Diffbot. engchi. i n U. v. Twitter infographic of Diffbot, 2012. 2 Diffbot (Aug 16, 2012). Diffbot’s New API Is a Decoder Ring for the Web. Retrieved from https://www.diffbot.com/company/news/20120816.jsp 2. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(13) 7 t. t Convolutional Neural Networks CNN. t. t. z. y. zt 8. 立. 7. ‧ y. t. n. al. sit. t. 8. ~. y. er. ‧ 國. 學. y. io. y. metadata w. Nat. z. 政治大. Cwh. w. engchi. i n U. v. v. 8. y. 8. y t. 7. 7. 8. 3. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(14) w. w. user-generated content, UGC. t. w. 7. 7 w. ~ w. y. 8 1-2. v. t7. 政治大. t. 立. ”. ‧ 國. 學. y. ~. ‧ sit. n. al. er. io. 7. t. y. Nat. 8. 7. y. Ch. engchi. i n U. v. w. w —. 8 t y. 8. v7. v. y. y 4. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(15) t 8. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. 1-2. 5. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(16) ”. 7. y v 8. y. t 7. 8. 政治大. io. w. v7. ”. 8. al. n. t. y. y. Nat. z. y. v. y. Ch 8. sit. ‧ 國. y. [3]. ‧. y. 學. 8t7. Mouzannar. w. 7. ~. w. i n U t. engchi. v. er. 立. v. y 8 Weng. ”. Lee(2011). Latent Dirichlet Allocation, LDA [4]. t ”. Abel. 8 (2012) ”. ”. [5]. 8 6. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(17) ~y 8. y. Mouzannar. y y Alam. y. [6]. 7. 7. y. 7. Peters. 8. Albuquerque(2015). 7. 2013. [7] z. y. 立. 8. Daly. 政治大 z. ‧ 國. 學 ‧ y. sit. n. al. er. io. 8. 8. Nat. ~. y. Ch. engchi. y y. Thomy(2016). i n U. v. t. y. 8. z. t. 8. 7. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(18) bag of words model. 1954. y. t “ t. Harris, Zellig[8] t5. 6y. 5. 6. —. 8. 7 y ~. 8. 立. 學. t. t. y. t. y. Nat. sit. t. n. al. er. io. 8. t. ‧. ‧ 國. y. 政治大. Ch. engchi. i n U. v. w. t. ~. 8 y z. w. 8. 8. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(19) Jieba. Stanford Word. Segmenter CKIP. t 8. computer vision. y. 7 t. —. 立. y 政治大 t. ‧ 國. t. ‧ Manish 7. n. er. io. [14]. al. y. y. Nikita. sit. feature. Nat. semantic. y. 學. 8. y. Ch. fingerprint. engchi. i n U. v. —. z. ~ 8. 3 Retrieved from https://github.com/fxsjy/jieba 4 Stanford Word Segmenter. Retrieved from https://nlp.stanford.edu/software/segmenter.shtm 9. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(20) Jain. 1995. Machine Vision t. [10]. —. labeling. —. — 2015. —. ImageNet. — —. t 2015. How-Old.net. t. z y. 2-1. 政治大 t. t. 立. w. t. 8. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. 2-1 How-Old.net How-Old.net. 5 Allison Linn Dec 10, 2015 . Retrieved from https://blogs.microsoft.com/next/2015/12/10/microsoftresearchers-win-imagenet-computer-vision-challenge/ 6 How old do I like? Retrieved from https://how-old.net/ 10. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(21) y. z. 8. t. y. 7 t. w. y t. 7. 立 ‧ 國. 8. sit. y. ‧. io. vertex. n. al. edge. er. Nat. graph 2-2. AntiDupl.NET. 學. t. Yermalayeu. 治政 t 32 大32 8. t. Ch. i n U. v. e n g c h i~. 8. t. t t. ~ ~ t. 8. y y. y directed. y weighted. 8. 7 Yermalayeu Ihar. AntiDupl.NET Description FAQ. Retrieved from http://antidupl.sourceforge.net/data/help/english/index.html?page=faq.html 11. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(22) t 7. 8. 政治大 Graph theory, Wikipedia. 2-2 t. 立. y. n. al. sit. 8. y. ~. y. t. er. io. —. ‧. ‧ 國. 學. ~ t. t. Nat. w. Ch. i n U w. engchi. v. ~ t. community. t ~. t. t. y. ~. t. 8. ~ y. y 8. label propagation. %. / ). 2007. [11]. t t. y. ~. t. 12. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(23) t. t Blondel. t t. 2008. [12]t. t. t. “. y t 8Pascal. [13]. 2005. w. t t. 8. y. t 8. 政治大. 立. indices y. ‧. recall. 7. sit. y. y. n. al. er. io. 8. t t. Nat. y. t. 學. precision. ~. ‧ 國. t. Ch. engchi. i n U. v. t evaluation. metrics. y Almeida [9]. Modularity. 0. t. 1. t. 2011. z t. 1. Coverage Converage. y. 8. ~. y 13. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(24) 7 8. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. 14. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(25) v ” 7. 8. ”. 7. v. t. 7. t. ”. 7. —. 8. 2016. 立. 921. 政 7治大. 7 8. ‧ Standard search API. Twitter. io. n. al. API. sit. y. Twitter. 7. er. 2016. w. Nat. AP. 學. ‧ 國. 8. Ch. n engchi U. iv. Twitter w. 3-1. 8. 5. 6. OR. Twitter. t. 8 1,224,586 27.06%. Twitter API 331,432. t. ~. Diffbot page classifier. 8 Twitter Standard search API https://developer.twitter.com/en/docs/tweets/search/api-reference/get-searchtweets 15. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(26) 2012. 35%. t. t. ”. 8. 3-1. TainanEarthquake 2/6/2016 4:38 UTC+08:00. 2/20/2016 22:19 UTC+08:00. 2/5/2016 15:30 UTC+08:00. 立. 政治大. y. sit. y. 84,423. 2,536. al. 8. n. 81,887. Ch. w. er. io. y. ‧. y. Nat. ”. 7. 學. 7. 331,432 w. t. 7. ‧ 國. w. 2/20/2016 21:59 UTC+08:00. engchi. i n U. v. 9 https://help.twitter.com/en/using-twitter/retweet-faqs t w w. y. t t. t 16. 8. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(27) 3-1 t. 立 ‧ 國. ‧ y. sit. n. al. er. io. 8. Nat. 2016. 學. ~. 政治大. Ch. engchi. i n U. v. 17. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(28) 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. 3-2. 3-3 34. 3-7 18. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(29) y t7 TZ. ~. 8. 政治大. 立. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. 3-3. Ch. engchi. i n U. v. w. 19. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(30) 3-4 TZ1 2016-02-06 06. 立. 政治大. ‧. ‧ 國. 學 sit. y. Nat. n. al. er. io. 3-5 TZ2 2016-02-07 18. Ch. engchi. 3-6 TZ3 2016-02-15 02. i n U. y. v. w. 20. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(31) 3-7 TZ4 2016-02-19 21. w. 政治大. 立. y 4. ‧. n. ayl. er. io. sit. y. Nat. 8. 學. y. ‧ 國. w. Ch. engchi. i n U. v. t y y. API. t ~. API. y. 8. 21. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(32) 3-8. 立. 政治大. ‧. ‧ 國. 學. wt. t. Nat. n. al. er. io. sit. y. 8. y. Ch. engchi U. v ni. y y. t 8. z. 1000. w. 8. 10 https://zh.wikipedia.org/wiki/%E4%B8%AD%E8%8F%AF%E6%B0%91%E5%9C%8B%E8%87%BA%E7 %81%A3%E5%9C%B0%E5%8D%80%E9%84%89%E9%8E%AE%E5%B8%82%E5%8D%80%E5%88% 97%E8%A1%A8 11 t http://elze.tanosii.net/d/kenmei.htm 22. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(33) 3-2. ~. 8. 1044636 4058. 1.. 420016. or. 2.. 19629. or. 3.. 968. or. and. 104090. or. 5144. 42062. or. or. 134553. or. 3000. or. or. 8322. 1. TaiwanEarthquake. 1711. or. 2. TainanEarthquake. 3717. or. 政167040治and 大Taiwan 119003 or Kaohsiung 2211 or Tainan 14296 or TW 32570 立 1,224,586. 7. w. ‧. 143,690. 學. ‧ 國. 3. Earthquake. n. al. er. io. sit. y. Nat —. 8. Ch. engchi. i n U. v. —. t. t —. w. y. 3-9 Google Vision API. 5Google6 Amazon Rekognition. 12. 7Microsoft Computer Vision API 5AWS6. 5MS6. 7. 7Watson Visual Recognition Service. or w 8 13 Google Vision API https://cloud.google.com/vision/ 14 Microsoft Computer Vision API https://azure.microsoft.com/zh-tw/services/cognitive-services/computervision/ 15 Amazon Rekognition https://aws.amazon.com/tw/rekognition/ 23 t. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(34) Clarifai. 5IBM6. 3-10. z. 7. Local. t. w. y. 7. 8. 政治大學. ‧ 國. 立. 3-9. ‧. Microsoft Computer Vision API. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. 3-10. 16 Watson Visual Recognition Service https://www.ibm.com/watson/services/visual-recognition/ 17 Clarifai https://clarifai.com/ 24. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(35) t. wt. 7. 8. t. w 8. 立 ‧ 國. 3-11. Mean square difference. 8. AntiDupl.NET. 學. wt. ‧. 2,265. 95%. Nat. io. sit. y. 8. n. al. er. UI. 政治大. Ch. engchi. i n U. v. 25. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(36) 立. 政治大. n. y. sit. UI. er. io. al. y. ‧. ‧ 國. 學. Nat. AntiDupl.NET. Ch. engchi. i n U. v. t 3-3 y Clarifai. IBM. 8. 3-3 26. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(37) Google. MS. AWS. IBM. Clarifai. 11017. 5121. 29979. 18482. 45300. 941. 417. 1125. 2260. 1563. 124. 406. 9. 0. 0. y 3-12. 3-16 y. t. 立. y t. 政治5 6大. —. y. “person”7“people”7“human”7“man”7“woman”7“adult” y. 30. ~. 8. y. sit. n. al. er. io. 3-4. “text”. y. ‧. coexistence. —. Nat. ~. ‧ 國. y. 學. z. t—. i ~ n C hy engchi U. v. 8. 27. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(38) 3-12 Google. 30. 3-13 MS. 立. 政治大. ‧. ‧ 國. 學 sit. n. 3-16 Clarifi. al. er. io 30. y. Nat 3-14 AWS. 30. Ch. i n U. 3-15 IBM. engchi. v. 30. 30 28. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(39) 3-4. t. w. z t—. 3-17. y certainty. 立. 學. —. y. w 政治大. ‧ 國. w. 3-18. 2,265 8. ‧. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. 3-17 29. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(40) 立. ‧. ‧ 國. 學 sit. io. al. 3-5. n. 3-21. y. Nat. 500. Ch. engchi U. er. 3-18. 政治大. v ni. 3-197. 3-207. 8. 30. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(41) 3-5. t. 7. 7. 20%. —. y. y v 20%. —. 立. 政 y 治 wv大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. 3-19. 31. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(42) 3-20. —. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. 3-21. engchi. i n U. v. 3-22. t 500. t. y. t t w t. y. z. 500. 8 32. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(43) 政治大. 立. ‧. ‧ 國. 學. 3-22. sit. n. al. Ch. 3. 8. engchi. 1989. 0.77. 18. 1. er. io. 2. y. Nat v. i n U. v. v. 0.91. (1989). 3-68. 8 33. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(44) 3-6. 1,2. 0.76. 1,3. 0.73. 2,3. 0.86 0.77 0.91 n*. ‧ 國. 500 95. —. 160. 32%. 19% w. 55. ‧. v. io. sit. y. Nat. 8. t. n. al. er. 11%. 38%. 立. *. 學. 190. 政治1 + 大n-1. Ch. engchi. i n U. v. 7. 3-23. 5 8. 34. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(45) 立. 3-23. ‧ 國. 學 AWS. io. n. al. Clarifi. v. IBM. t. ~. er. Clarifi. Google. 3-8. sit. y. Nat. Google. 3-7. t. ‧. w. 政治大. Ch. w. i n IBM engchi U. MS. v. —. MS. v MS. z. AWS. 8 AWS. 8. 35. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(46) Google. MS. AWS. IBM. Clarifi. 63.64. 65.46. 78.18. 54.54. 69.09. 64.21. 0. 75.79. 67.37. 50.52. 44.21. 0. 80.53. 62.11. 82.63. Google. MS. AWS. IBM. Clarifi. 22.27. 23.53. 17.72 治政 0 17.69 大. 26.51. 16.52. 5.77. 22.21. 0. 8.65. 18.39. 10.13. 立. ‧ 國. t. ‧. 7. 2.35. 學. 14.22. sit. n. a”l. er. io 8. y. Nat. —. y. Ch. i n U. v. e n g c h it. 8. 36. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(47) v t. 8. 7 ~ t. y. t w. 8. 政治大. v. 509,256. 65.37%. ‧. 5. 27.06%. 學. ‧ 國. 立. 6. 8. sit 5. n. al. Ch. Twitter API. 6. ” engchi U. er. io. y. y. Nat. t. 5. v ni. w 6. 5. 6. z. Twitter. y 4-1. t t. ~ 8. 37. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(48) 4-1. 1. 11. 12. 1. 2. 3. 1. 4. 5. 33.33%. 立. 60%. 政治大. t. Twitter API. y. ‧ 國. 學. wt. ‧. 8. Twitter w. t. n. al. er. io. sit. Nat. 8. y. 8. y. 143,690 t. Ch. 65.37%. engchi. i n U. v. 10.19%. w. t 18.09. 4.31. y. t. w. t y 8. 38. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(49) AWS. ~. t. y 8. AWS. v. y. t. brute force t. w. t. Modularity. y. t 8. 立 ‧ 國. 2,265. 1,125. t. y. n. er. io. 8. al. 7. y. —. sit. 7. —. ‧. —. Nat. —. 29,979. 學. y. 政治大. Ch. engchi. y. i n U. y. v. t. global. local. 8. 4-1 50-100 ~ 57.91%. AWS 60. 75.2% 8.21%. 70. 39. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(50) 86.31%. 70.37% 5.17% 60. 60. 57.91%. y 24.8%. 8.21%. y 60. ~. 91.79%. 8. 政治大. 立. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. 4-1. wt IBM 5. 6. y 5. — 6. y. —. w t. 8. t. Santorini. 40. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(51) [15] González. 2005. [16]. 7. 7. 7. 立. 8. 政治大學 ‧. ‧ 國 io. sit. y. Nat. n. al. er. 1990. Ch. engchi. i n U. v. 4-2. 41. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(52) 4-3. 立. y. ‧ 國. 學. y. sit. n v. 7. er. io. al. 7. 7. v. Ch. 7. engchi U. v. t. y. 7. v ni. 0.1. v. igraph walktrap. 8. y. Nat. 4-4 7. t. ‧. Modularity. 政治大. louvain7label_prop. walktrap. 5. 15. 710. 8. y. Modularity 4-2. wy. 7. louvain ”. 19 igraph. Modularity. 8. http://igraph.org/r/ 42. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(53) 4-2 y. Modularity 7. 7. 7. 立. 10 7. 7Modularity. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. 43. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(54) 政治大. 立. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. wt. Ch. engchi. y. i n U. v. y Modularity. y w 2.2. w. louvain. zy 8. 8 y. 5%. 4-63. 4-13. 8. 44. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(55) 4-3. 3. 3. ~. ~ RT. 1% t. RT. RT. 1%. 8. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. 4-5. 45. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(56) 4-3. person7human7 — —. 3. people7fireman7boat. 55. 73. 3613. 1165. 18. 9. rubble7demolition. 52. 101. 2822. 776. 16. 42. 61. 3627. 4085. 9. 1205. 553. 11. 189. 26. 0. 926. 328. 7. 1193. 1214. 8. 2961. 2171. 8. downtown7city7 road7intersection7 28. aerial view. 治政大 collage7brochure7 立 paper 36 38 poster7flyer7. 學. ‧ 國. 10. text7label7book7. 2. 27. 29. map7atlas. 23. 31. 22. 47. y. Nat. credit card7letter. ‧. 25. n. al. Ch. plant7potted7plant7 fence7hedge. 4. —. 4-6. 3. engchi 22. i n U. 4-7. —. er. io. motorcycle7stage7car. 12. sit. vehicle7motor7. v. 32. 9. —. 46. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(57) 4-8. 28. 4-9. 立. 10. 政治大. ‧. ‧ 國. 學. n. al. 4-12. Ch. engchi. 12. sit. 4-11. 2. er. io. 25. y. Nat 4-10. i n U. 4-13. v. 4. —. 47. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(58) 7 y. y. 7. —. t w. t LOGO. 4-16. w. “plant”. 4-15. 4. 8. z. 政治大. w. “plant”. 學. ‧ 國. 107 t. 立. — ”. “plant” 0.18%. —. — 4-14,. y. 8. w. ‧. 8. n. er. io. sit. y. Nat. al. Ch. engchi. 4-14. —. i n U. v. y. 48. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(59) 4-15. y. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. 4-16. Ch. engchi. i n U. v. —. 49. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(60) 3.3.3. ~. 4-4 9. 28. —. —. 10 w 8. 4-4. 政治大. 立 9. 52. ( 78.85% /41) ( 21.15% /11). ( 75% /27), ( 13.89% /5), ( 11.11% /4). y. 36. n. al. 42. ( 95.24% /40) — ( 2.38% /1) ( 2.38% /1). sit. io. 10. Nat. 28. —. er. ‧ 國. 55. 25. —. ‧. —. 3. 學. —. ( 63.64% /35) ( 32.73% /18) ( 3.64% /2) —. C 27 h. 2. 23. 12. 22. 4. 22. i n U. v. ( 51.85% /14). engchi. ( 95.65% /22),. —. ( 44.44% /12), ( 3.7% /1) ( 4.35% /1). ( 68.18% /15), ( 18.18% /4), ( 13.64% /3) ( 59.09% /13) ( 36.36% /8), — ( 4.55% /1). 50. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(61) t. 10. t ID. tt. 7 y. y. 8. ID. z 8 2007. y. 立. t 8. ‧. ‧ 國. 學. io. sit. y. Nat. n. al. er. 4-5. 7 7 治政大. [17]. Ch. engchi. i n U. v. 51. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(62) media. 7. 7 8. organization. 7. 7 8. bot. bot 8. user. t. 立. v. sit t. Ch. 4-7. v. er. al. 4-6. y. 2. n. 12. w. ‧. 10. io. 10. y 8. Nat. 1. 政治大. 學. ‧ 國. t. engchi U —. v ni. 25 972872 wt t. 8. 52. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(63) 4-6. 3. media(6), user(4). 9. media(9), user(1). 28. media(8), user(2). 10. user(9), media(1). 25. user(10). 2. media(7), news(1), organization(1), bot(1). 4. user(6), media(3), organization(1). — —. 政治大 media(7), user(3). 12. —. 立 ‧ 國. 學. uid. 2. USGS. 2. China Xinhua News. XHNews. media. 2. Newsweek. Newsweek. media. 2. KTLA. KTLA. media. 2. TIMES NOW. TimesNow. media. io. RT_com. 2. !. 2. Times of India. 2. Australian News Net. Ch. USGS. organizati on. engchi. earthquake_al. bot. l timesofindia AusNewsNet work. media media. 282. 186. RT. 269. 85. v. 164. 55. 142. 54. 81. 24. 46. 39. 42. 34. 40. 21. 30. 19. 16. 1. t. media. n. al. AccuWeather. y. RT. her. news. sit. 2. Nat. AccuWeather. ‧. breakingweat. 2. er. uname. 9. i n U. t. TIMES NOW ! Times of India. 53. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(64) 3. KINBRICKS NOW. 3. kinbricksnow. user. 1,065. 133. suizou. media. 554. 89. suizou. 3. _pB R. 17. 10ntyan. user. 479. 67. 3. H(a. ). Tempatmata. user. 452. 86. 3. beh _. media. 312. 98. 317. 95. 239. 41. 123. 18. 117. 30. 110. 19. 2,016. 688. 481. 280. 284. 74. 149. 36. 127. 18. media. 73. 104. user. 47. 8. 34. 24. a r. kinkakuji09. 立 joe_i. Joe ISHIYAMA. a r. focustaiwanja pa. r. _. gn o. r. io. 4 4. Joint Cyclone Center. 4. ABS-CBN News. 4. Russian Market. 4. user. yuki4x3. user. JointCyclone ABSCBNNe ws. organizati on. russian_mark et ANCALERT. Channel. S. Teabar.com. _. a ltaehitoto user iv n C h e n media nhk_seikatsu gchi U. ABS-CBN News. 4 4. Maechan0502. n. t. media. y. =B R @. user. ‧. e. media. 治政 media 大. Srir. 4 4. suizou. Nat. 4. suizou. _. media. 學. 3. pa. r. sit. 3. focustaiwanja. er. e. 3 3. n. ‧ 國. 3. HuffPostJapa. media. _. LazyWorkz. user. 31. 9. MAHAMOS. user. 17. 30. 54. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(65) A T. 9. iza_edit. media. T. 717. 241. CNN_. 385. 303. 9. CNN Breaking News. cnnbrk. media. 9. CNN. CNN. media. CNN. 269. 248. 9. NBC News. NBCNews. media. NBC News. 153. 92. livedoornews. media. 128. 61. 95. 55. 96. 57. BBC News Japan. 90. 29. 政治 BBC bbcnewsjapan media 大News Japan 立 hongkong_bl. 85. 35. 76. 23. 164. 45. 118. 23. 116. 33. 112. 25. 107. 17. 83. 13. 66. 9. 9. T. 9. CNN International. cnni. media. 9. China Xinhua News. XHNews. media. 9. BBC News Japan. bbcnewsjapan. media. taehitoto. user. ". #. HenryLauFansPage. Nat. 10. D. 10. user. tsubuchai. user. henrylaufp. user. aoKaeru. user. media BNO News aBNONews iv l linsbar user n Ch engchi U. n. BNO News. io. 10. zakichu. 10 10. ‧. 10. 學. I. 10. user. og. $%. er. 10. BBC News Japan. ‧ 國. 9. y. t. 9. CNN_. sit. Sa r. tda9491. user. 64. 15. 10. chako. joj_chako. user. 28. 5. 10. Mickey Fulp. mercenarygeo. user. 19. 2. 12. Reuters Top News. Reuters. media. 1,123. 282. 104. 38. 90. 26. 79. 53. 66. 20. 12. a r e. r. focustaiwanja pa. media. 12. The Boston Globe. BostonGlobe. media. 12. Times of India. timesofindia. media. 12. The Telegraph. Telegraph. media. _ _. Times of India. 55. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(66) 12. Reuters World. ReutersWorld. media. _. 46. 8. 12. News18. CNNnews18. media. CNN_. 44. 26. 12. Ahmer Alavi. ahmeralavi. user. 33. 4. napudazo. user. 21. 4. doridoriroom. user. 21. 8. babi_sano. user. 22. 7. ikon_infinity. user. 19. 8. 13. 11. 11. 3. 10. 4. 10. 13. 10. 2. 7. 14. 7. 3. 4. 3. 3,691. 614. 343. 74. K. 12. m ★. >. CH Forever With iKON '. 政 user治大 e. shimizu_kaed. 25. user. Anthony Lomax ( )* ALomaxNet. user. 25. 25. 제노.. 28 28. 28 28. Isseki3. user. joj_chako. user. a leethejeno user iv l C n he gchi U nhk_seikatsu n media CGTNOfficia. CGTN. l. a r e. user. n. chako. myamuis1. t. io. 25. user. Nat. Isseki Nagae/. 4NOvicious. focustaiwanja. a r. media. media. _. 272. 58. Sankei_news. media. a r. 113. 29. 89. 39. 68. 29. 45. 19. pa. r. 28. N 7/28.29. kohaku5237. user. 28. The Independent. Independent. user. 28. Taiwan TW_nextmed News. ‧. bass. 25. 25. R@ MA. 學. 25. 立 satoshickbkr. R. ‧ 國. 25. y. 25. m. sit. 25. ni a by. er. 12. ia. media. “ Taiwan News. 56. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(67) 28. ITV News. itvnews. media. ITV News. 43. 14. 37. 32. 30. 10. Ruptly TV 28. Ruptly. 28. Wp Tp. Ruptly. media. Toyokeizai. media. RT. Wp Tp. ~ 1% 4-17. 1%. 186. 立. 8. 2. 治 ” 政大. 122. 學 ‧. ‧ 國 io. sit. y. Nat. n. al. er. 4-7. 124. Ch. engchi. i n U. v. 4-17. 57. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(68) 4-8. media. 24. ) 01. 28. 12. 9. 12. 9. 7. 54. 53. 2. . ( -. 14. 1. 立. .0 .. , ) ,/ . ,. ‧. . .. .. 7. io. sit. 76. y. Nat. user. 8. 政治大學. bot. ‧ 國. ion. 1. n. al. er. organizat. 4-7. Ch. engchi. i n U. v. t. 7. w w. 8. t. t. t 2016. zt. y. 7. 7. 7. 58. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(69) t. ~. 5. 6. y. 5. 6 8. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. 59. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(70) 2016 Twitter. t Twitter. t. 8. t. 7 10.19%. y. 政治大. 65.37%. w. 立 ‧ 國. ~. ”. y. sit. al. n. Rekognition. Twitter. er. io. w—. 8. ‧. 8. Nat. tt. Twitter API. t. 學. Twitter. y. Ch. engchi. i n U y. v. Amazon ~. 7. ”. 8. y. — — t. y. —. 60. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(71) —. y. y. ~. 8. t t. ”. t t. t. y. t. t. z t. y. y. 立 ‧ 國. 8. y. y. sit. t t. n. al. er. io. 7. ‧. —. t. 學. t. Nat. y. 政治大. 7CNN7The New York Times7. Ch. engchi. 7. i n U. v. 7BBC. RT. 7 y. t ts. 8. 61. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(72) l [1] Imran, M., Castillo, C., Diaz, F., & Vieweg, S. 2015 . Processing social media messages in mass emergency: A survey. ACM Computing Surveys CSUR , 47 4 67.. ,. [2] Daly, S., & Thom, J. A. 2016, May . Mining and Classifying Image Posts on Social Media to Analyse Fires. In ISCRAM. [3] Mouzannar, H., Rizk, Y., & Awad, M. Damage Identification in Social Media Posts using Multimodal Deep Learning. [4]Weng, J., & Lee, B. S. (2011). Event detection in twitter. ICWSM, 11, 401-408. [5]Abel, F., Hauff, C., Houben, G. J., Stronkman, R., & Tao, K. (2012, June). Semantics+ filtering+ search= twitcident. exploring information in social web streams. In Proceedings of the 23rd ACM conference on Hypertext and social media (pp. 285-294). ACM.. 政治大 [6]Alam, F., Imran, M., & Ofli, F. (2017, July). Image4act: Online social media image 立 processing for disaster response. In Proceedings of the 2017 IEEE/ACM International ‧ 國. 學. Conference on Advances in Social Networks Analysis and Mining 2017 (pp. 601-604). ACM.. . Distributional Structure. Word, 10. 2/3 , 146–162. sit. y. 1954. Nat. [8] Harris, Z. S.. ‧. [7]Peters, R., & de Albuquerque, J. P. (2015). Investigating images as indicators for relevant social media messages in disaster management. In ISCRAM.. n. al. er. io. [9] Upadhyaya, N., & Dixit, M. 2016 . A Review: Relating Low Level Features to High Level Semantics in CBIR. International Journal of Signal Processing, Image Processing and Pattern Recognition, 9 3 , 433-444.. Ch. engchi. [10] Jain, R., Kasturi, R., & Schunck, B. G. York: McGraw-Hill.. 1995. i n U. v. . Machine vision. Vol. 5. . New. [11]Raghavan, U. N., Albert, R., & Kumara, S. 2007 . Near linear time algorithm to detect community structures in large-scale networks. Physical review E, 76 3 , 036106. label [12]Blondel, V. D., Guillaume, J. L., Lambiotte, R., & Lefebvre, E. 2008 . Fast unfolding of communities in large networks. Journal of statistical mechanics: theory and experiment, 2008 10 , P10008. [13] Pons, Pascal, and Matthieu Latapy. "Computing communities in large networks using random walks." ISCIS. Vol. 3733. 2005.. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(73) [14] Almeida, Hélio, et al. "Is there a best quality metric for graph clusters?." Joint European Conference on Machine Learning and Knowledge Discovery in Databases. Springer, Berlin, Heidelberg, 2011. [15]Santorini, B. 1990 . Part-of-speech tagging guidelines for the Penn Treebank Project 3rd revision . Technical Reports CIS , 570. [16]González, F.A., Gelbukh, A.F., & Jiménez, S. 2015 . Soft Cardinality in Semantic Text Processing: Experience of the SemEval International Competitions. Polibits, 51, 6372. y &. 2017 41. 立. Twitter. d 81-117. 政治大學 ‧. ‧ 國 io. sit. y. Nat. n. al. er. [17]. Ch. engchi. i n U. v. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(74) 自定義名稱. 系統名稱. 立. 跟隨者. 人工判斷類別. 政治大. ‧ 國. 學. 地震速報. ‧ sit. y. Nat. n. al. er. io. NHKニュース糸井重里. tenki.jp地震情報. Ch. engchi. i n U. v. 西川貴教ふなっしーそらるＮＨＫ生活・防災. 楽天イーグルス. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(75) !"#$%& ライブドアニュース特務機関NERV. Yahoo!ニュースシドマオ新田恵海. 立. 石平太郎. 政治大. ‧ 國. 學. 高須克弥. ‧. 林雄介・新刊(宗教で得する人・損する人). sit. y. Nat. io. al. n. ハフポスト日本版. er. ZIP! 日テレ. 産経ニュース. Ch. engchi. i n U. v. BROS.（福山雅治公式ファンクラブ）緊急地震速報bot（β）. 佐藤正久鈴木拡樹. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(76) 佐藤拓也. 松本梨香 47NEWS 速報地震速報. 駒崎弘樹＠障害児保育スタッフ募集中山際澄夫 Yahoo! JAPAN（ヤフー）. 立. 政治大. でじたろう@ニトロプラス. ‧ 國. 學. 地震うさぎ. ‧. 地震速報. Nat. sit. n. er. io. al. y. ☆Chris*台湾人☆. 桜井誠. Ch. engchi. i n U. v. H(ニャ民党) TBSテレビ「あさチャン！」あそうかも。毎日新聞写真部銀座ウエスト KINBRICKS NOW（高口康太）. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(77) 台湾人不破雷蔵イザ！編集部 Re:file（リファイル）日本赤十字社台湾ニュース＠中央社フォーカス台湾愛国魂一本木蛮ｺｺｺﾐまめしばコ連載中 ♥. 政治大 NewsDigest ニュース・災害速報立 ‧ 國. 學. 安達裕章. ‧. ゆる系速報・ぽん太くん. y. Nat. n. er. io. al. sit. さとる＠金沢暮らし【飯と音楽】. Ch. 吉田凜音STAFF4/25アルバム. engchi. i n U. v. 一青妙. まえちゃん @ワーホリ電子書籍発売中吉良青劉 Yahoo!基金夜ツイ. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(78) ブラスコウ／秋友克也朝比 ☆テッドのブラック雑学☆ 武藤義明 '. 富田恭敏九条先輩☆ひろし☆戦争法絶対廃棄！傘下逆. 政治大. 立. ‧ 國. 學 ‧. 由希. sit. er. al. n. T. io. 炭. y. Nat. トンちゃん（17）. うんちく@例大祭 F-51b 迅雷数奇. Ch. engchi. i n U. v. (. ニュース速報24 きなこもち ♨ ねこやまだ. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(79) オフマト ECOシンガー (堺の徳永英明)(堺のつるの剛士) ほくつゆり東雲情報局. テイゴン＠勇者部. 立. 學. ‧ 國. 如月弥生. 政治大. 言いそびれたが私はミテイ. ‧. 三角頭. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. DOI:10.6814/THE.NCCU.CS.004.2018.B02.

(80)