隨著互聯網技術的高速發展，社群媒體成為了人們獲取資訊的主要途徑。然而，由於缺乏及時的謠言偵測系統，現在的社群媒體已經淪陷為謠言的聚集地。這些充斥在社群媒體中的謠言會威脅人民的身體健康，危害經濟的發展，甚至影響國家的安定繁榮。因此建立一個自動謊言偵測系統迫在眉睫。但是現階段的自動謠言偵測技術由於可用文本太少，資料擷取不充分，不具備模擬資訊在網路傳播，不具備有可解析性等問題，不能被運用於實際。為了解決現階段自動謠言偵測技術存在的問題，本文將使用人工智慧的自然語言處理的技術來對自動謠言偵測任務進行研究。
為了解決上述自動謠言偵測任務面臨的問題，本文提出了四種基於人工智慧自然語言處理的自動謠言偵測模型。這四種模型均具備有良好的謠言偵測性能。第一個模型是帶衰減因數雙反向傳播層級結構模型，該模型提出以層級結構的模式來處理謠言，同時結合多任務的思想，為層級結構接入了雙反向層和衰減因數，即提升了謠言偵測的精確度，還能緩解梯度消失的問題，減少模型訓練時間。第二個模型是可處理對話結構資訊的謠言偵測模型，該模型是首個提出使用對話結構來代替線性結構的模型，也是首個具備模擬謠言在社群媒體中的傳播的模型。它的提出進一步提升了當前謠言偵測的準確度。第三個模型是利用同質性和對話結構的謠言偵測模型，該模型首創的使用了朋友關係網路的同質性，結合對話結構和用戶行為模式來進行謠言偵測。在同質性的幫助下，該模型還解決了現階段早期謠言偵測精確度低的問題。第四個模型是提供證據的可解釋的謠言偵測模型。這一個模型彌補了現階段謠言偵測模型不具備有解釋性的技術缺陷。實現了謠言偵測與闢謠的雙重功能。
本研究還進一步探索分析了這四個模型的性能。本文使用了兩個著名謠言偵測數據集來測試其模型性能。實驗結果表明，這四個模型模型提升了現階段謊言偵測的精確度，與現階段的通用模型相比，提升了30%以上的偵測精確度。同時這四個模型還具有良好的早期謠言偵測能力，可以依據極少量的資訊進行快速準確的早期謠言偵測。且模型都具有較好的相容性和穩健性，稍作修改還可以運用新聞偵測，郵件偵測等領域。
關鍵字：謠言偵測、圖神經網路、對話結構、可解釋性、證據檢索、同質性

Due to rapid development of internet technology, social media has become a primary information source for many people. However, social media has become a breeding ground for spreading rumors due to the lack of an automatic rumor detection system. These rumors endanger people's health, impede economic development, and jeopardize a country's stability and prosperity. As a consequence, it is extremely urgent to establish an automatic rumor detection system. However, current automatic rumor detection technologies cannot be applied to real-world social media platforms due to insufficient data extraction, interpretability, and information transmission simulating ability.
This dissertation proposes four deep learning-based rumor detection models exhibiting good rumor detection performance to address the aforementioned issues. The first model is a rumor detection model that employs a multi-loss hierarchical BiLSTM with an attenuation factor. This model processes rumors in a hierarchical structure in conjunction with a multi-task mechanism. The model is divided into two BiLSTM modules and can extract deep information from limited quantities of text. Each module includes a loss function that assists in learning bilateral rumor features and reducing training time. An attenuation factor is added at the post-level module to increase model’s detection accuracy. The second model is a rumor detection model that can process conversational structure information. This model is the first to propose using conversation structure rather than linear structure, and it is also the first to simulate rumor propagation in social media. It further improves the accuracy of current rumor detection. The third model is a rumor detection model based on homogeneity and conversation structure. This model is the first to detect rumors by employing homogeneity of friends network and combining conversation structure and user behavior pattern. With the help of homogeneity, the model also addresses the issue of low accuracy of rumor detection in the early stage. The fourth model is an interpretable, evidence-based rumor detection model. This model fills in the explanation gap left by existing rumor detection models, allowing rumor detection models to illustrate why a suspicious statement is a rumor.
This study examines and analyzes the performance of these four models in detail. For each model, two real-world rumor detection datasets are used for evaluation. Compared with current baseline rumor detection models and state-of-the-art models, experimental results indicate that these four models achieve good rumor detection performance and outperform other baselines in improving the detection accuracy by over 30%. Additionally, the proposed models also have perfect early rumor detection performance. Experimental results indicate that the models can quickly and accurately detect rumors based on a limited amount of information at the initial stage of rumor propagation. Additionally, the proposed models have good compatibility and robustness, allowing them to be employed not only for rumor detection but also for news detection, email detection, and other fields with minor modifications.
Keywords: Rumor detection, Graph neural network, Convolution structure, Evidence retrieve, Interpretability, Homogeneity

[1] N. Diakopoulos, M. De Choudhury, and M. Naaman, “Finding and assessing social media information sources in the context of journalism,” in Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, 2012, pp. 2451–2460.
[2] P. Tolmie et al., “Supporting the use of user generated content in journalistic practice,” in Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems, 2017, vol. 2017-May, pp. 3632–3644.
[3] R. Procter, F. Vis, and A. Voss, “Reading the riots on Twitter: Methodological innovation for the analysis of big data,” Int. J. Soc. Res. Methodol., vol. 16, no. 3, pp. 197–214, 2013.
[4] H. Webb et al., “Digital wildfires: Propagation, verification, regulation, and responsible innovation,” ACM Trans. Inf. Syst., vol. 34, no. 3, 2016.
[5] P. Tolmie, R. Procter, M. Rouncefield, M. Liakata, and A. Zubiaga, “Microblog Analysis as a Program of Work,” ACM Trans. Soc. Comput., vol. 1, no. 1, pp. 1–40, Jan. 2018.
[6] R. Procter, J. Crump, S. Karstedt, A. Voss, and M. Cantijoch, “Reading the riots: what were the police doing on Twitter?,” Polic. Soc., vol. 23, no. 4, pp. 413–436, 2013.
[7] H. Li and Y. Sakamoto, “Computing the veracity of information through crowds: A method for reducing the spread of false messages on social media,” Proc. Annu. Hawaii Int. Conf. Syst. Sci., vol. 2015-March, pp. 2003–2012, 2015.
[8] J. Cao, J. Guo, X. Li, Z. Jin, H. Guo, and J. Li, “Automatic Rumor Detection on Microblogs : A Survey,” arXiv Prepr., p. arXiv:1807.03505.
[9] Q. Li, Q. Zhang, and L. Si, “Rumor Detection By Exploiting User Credibility Information , Attention and Multi-task Learning,” in Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, 2019, pp. 1173–1179.
[10] O. Ajao, D. Bhowmik, and S. Zargari, “Fake News Identification on Twitter with Hybrid CNN and RNN Models,” in Proceedings of the 9th International Conference on Social Media and Society, 2018, pp. 226–230.
[11] Z. Zhao, P. Resnick, and Q. Mei, “Enquiring Minds: Early Detection of Rumors in Social Media from Enquiry Posts,” in Proceedings of the 24th International Conference on World Wide Web, 2015, pp. 1395–1405.
[12] P. Metaxas and S. Finn, “Using TwitterTrails . com to Investigate Rumor Propagation,” in Proceedings of the 18th ACM Conference Companion on Computer Supported Cooperative Work & Social ComputingFebruary, 2015, no. February, pp. 69–72.
[13] G. Liang, W. He, C. Xu, L. Chen, and J. Zeng, “Rumor Identification in Microblogging Systems Based on Users ’ Behavior,” IEEE Trans. Comput. Soc. Syst., vol. 2, pp. 99–108, 2016.
[14] N. Difonzo and P. Bordia, “Rumor , Gossip and Urban Legends,” Diogenes, vol. 54, no. 1, pp. 19–35, 2007.
[15] P. Donovan, “How Idle is Idle Talk? One Hundred Years of Rumor Research,” Diogenes, vol. 54, no. 1, pp. 59–82, 2007.
[16] C. Castillo, M. Mendoza, and B. Poblete, “Information credibility on Twitter,” in Proceedings of the 20th International Conference Companion on World Wide Web, 2011, pp. 675–684.
[17] F. Yang, Y. Liu, X. Yu, and M. Yang, “Automatic detection of rumor on Sina Weibo,” in Proceedings of the ACM SIGKDD Workshop on Mining Data Semantics - MDS ’12, 2012, pp. 1–7.
[18] K. Wu, S. Yang, and K. Q. Zhu, “False rumors detection on Sina Weibo by propagation structures,” in Proceedings of the IEEE 31st International Conference on Data Engineering, 2015, vol. 2015-May, pp. 651–662.
[19] S. Kwon, M. Cha, and K. Jung, “Rumor Detection over Varying Time Windows,” PLoS One, vol. 12, no. 1, p. e0168344, Jan. 2017.
[20] J. Ma, W. Gao, Z. Wei, Y. Lu, and K.-F. Wong, “Detect Rumors Using Time Series of Social Context Information on Microblogging Websites,” in Proceedings of the 24th ACM International on Conference on Information and Knowledge Management, 2015, pp. 1751–1754.
[21] B. Rath, W. Gao, J. Ma, and J. Srivastava, “From Retweet to Believability: Utilizing Trust to Identify Rumor Spreaders on Twitter,” in Proceedings of the 2017 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining, 2017, pp. 179–186.
[22] J. Ma et al., “Detecting Rumors from Microblogs with Recurrent Neural Networks,” in Proceedings of the 25th International Joint Conference on Artificial Intelligence, 2016, pp. 3818–3824.
[23] K. Xu, S. Jegelka, W. Hu, and J. Leskovec, “How powerful are graph neural networks?,” in Proceedings of the 7th International Conference on Learning Representations, ICLR, 2019, pp. 1–17.
[24] T. Chen, L. Wu, X. Li, J. Zhang, H. Yin, and Y. Wang, “Call Attention to Rumors: Deep Attention Based Recurrent Neural Networks for Early Rumor Detection,” in Trends and Applications in Knowledge Discovery and Data Mining. PAKDD, 2018, p. 11154.
[25] F. Yu, Q. Liu, S. Wu, L. Wang, and T. Tan, “A Convolutional Approach for Misinformation Identification,” in Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence, 2017, pp. 3901–3907.
[26] J. Ma, W. Gao, and K.-F. Wong, “Rumor Detection on Twitter with Tree-structured Recursive Neural Networks,” in Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2018, pp. 1980–1989.
[27] F. Monti, F. Frasca, D. Eynard, D. Mannion, and M. M. Bronstein, “Fake News Detection on Social Media using Geometric Deep Learning,” arXiv Prepr., p. arXiv:1902.06673, 2019.
[28] N. T. Tam, “Graph-based Rumour Detection for Social Media,” Comput. Sci., 2019.
[29] H. Guo, J. Cao, Y. Zhang, J. Guo, and J. Li, “Rumor Detection with Hierarchical Social Attention Network,” in Proceedings of the 27th ACM International Conference on Information and Knowledge Management,CIKM, 2018, pp. 943–951.
[30] Y. Liu and Y. F. B. Wu, “Early detection of fake news on social media through propagation path classification with recurrent and convolutional networks,” in Proceedings of the 32nd AAAI Conference on Artificial Intelligence, 2018, no. December, pp. 354–361.
[31] Y. Yang, L. Zheng, J. Zhang, Q. Cui, Z. Li, and P. S. Yu, “TI-CNN: Convolutional Neural Networks for Fake News Detection,” arXiv Prepr., p. arXiv:1806.00749, 2018.
[32] N. Ruchansky, S. Seo, and Y. Liu, “CSI: A hybrid deep model for fake news detection,” Int. Conf. Inf. Knowl. Manag. Proc., vol. Part F1318, pp. 797–806, 2017.
[33] Y.-J. Lu and C.-T. Li, “GCAN: Graph-aware Co-Attention Networks for Explainable Fake News Detection on Social Media,” in Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics,ACL, 2020, pp. 505–514.
[34] A. Hanselowski et al., “UKP-Athene: Multi-Sentence Textual Entailment for Claim Verification,” in Proceedings of the First Workshop on Fact Extraction and VERification (FEVER), 2019, pp. 103–108.
[35] J. Luken, N. Jiang, and M.-C. de Marneffe, “QED: A fact verification system for the FEVER shared task,” in Proceedings of the First Workshop on Fact Extraction and VERification (FEVER), 2019, pp. 156–160.
[36] Y. Nie, H. Chen, and M. Bansal, “Combining fact extraction and verification with neural semantic matching networks,” in Proceedings of the Thirty-Third AAAI Conference on Artificial Intelligence, 2019, pp. 6859–6866.
[37] C. Malon, “Team Papelo: Transformer Networks at FEVER,” in Proceedings of the First Workshop on Fact Extraction and VERification (FEVER), 2019, pp. 109–113.
[38] J. Zhou et al., “Gear: Graph-based evidence aggregating and reasoning for fact verification,” in Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference, 2020, pp. 892–901.
[39] Z. Liu, C. Xiong, M. Sun, and Z. Liu, “Fine-grained Fact Verification with Kernel Graph Attention Network,” in Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics,ACL, 2019, pp. 4171–4186.
[40] W. Zhong et al., “Reasoning Over Semantic-Level Graph for Fact Checking,” in Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics,ACL, 2020, no. 2018, pp. 6170–6180.
[41] T. Yoneda, J. Mitchell, J. Welbl, P. Stenetorp, and S. Riedel, “Four Factor Framework For Fact Finding (HexaF),” pp. 97–102, 2018.
[42] A. Zubiaga, A. Aker, K. Bontcheva, M. Liakata, and R. Procter, “Detection and Resolution of Rumours in Social Media: A Survey,” ACM Comput. Surv., vol. 51, no. 2, pp. 1–36, Jun. 2018.
[43] J. Ma, W. Gao, and K.-F. Wong, “Detect Rumor and Stance Jointly by Neural Multi-task Learning,” in Proceedings of the Web Conference,WWW, 2018, pp. 585–593.
[44] Y. Kim, “Convolutional Neural Networks for Sentence Classification,” in Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing ,EMNLP, 2014, vol. 218, pp. 1746–1751.
[45] A. Zubiaga, M. Liakata, and R. Procter, “Exploiting Context for Rumour Detection in Social Media,” in Lecture Notes in Computer Science, vol. 10539 LNCS, Cham, 2017, pp. 109–123.
[46] A. Zubiaga, M. Liakata, and R. Procter, “Learning Reporting Dynamics during Breaking News for Rumour Detection in Social Media,” arXiv Prepr., p. arXiv:1610.07363, Oct. 2016.
[47] X. Liu, A. Nourbakhsh, Q. Li, R. Fang, and S. Shah, “Real-time Rumor Debunking on Twitter,” in Proceedings of the 24th ACM International on Conference on Information and Knowledge Management,CIKM, 2015, pp. 1867–1870.
[48] M. Lukasik, T. Cohn, and K. Bontcheva, “Classifying Tweet Level Judgements of Rumours in Social Media,” in Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, 2015, no. September, pp. 2590–2595.
[49] M. Lukasik, P. K. Srijith, D. Vu, K. Bontcheva, A. Zubiaga, and T. Cohn, “Hawkes Processes for Continuous Time Sequence Classification: an Application to Rumour Stance Classification in Twitter,” in Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), 2016, pp. 393–398.
[50] M. Mendoza, B. Poblete, and C. Castillo, “Twitter under crisis,” in Proceedings of the First Workshop on Social Media Analytics ,SOMA, 2010, pp. 71–79.
[51] E. Rosengren, D. R. Radev, Q. Mei, and A. Arbor, “Rumor has it: identifying misinformation in microblogs,” in Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing, 2011, pp. 1589–1599.
[52] Y.-C. Chen, Z.-Y. Liu, and H.-Y. Kao, “IKM at SemEval-2017 Task 8: Convolutional Neural Networks for stance detection and rumor verification,” in Proceedings of the 11th International Workshop on Semantic Evaluation (SemEval-2017), 2017, no. 2011, pp. 465–469.
[53] I. Augenstein, T. Rocktäschel, A. Vlachos, and K. Bontcheva, “Stance Detection with Bidirectional Conditional Encoding,” in Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, 2016, vol. 35, no. 6, pp. 876–885.
[54] K. Zhou, C. Shu, B. Li, and J. H. Lau, “Early Rumour Detection,” in Proceedings of the 2019 Conference of the North, 2019, pp. 1614–1623.
[55] R. Belkaroui, R. Faiz, and A. Elkhlifi, “Conversation Analysis on Social Networking Sites,” in 2014 Tenth International Conference on Signal-Image Technology and Internet-Based Systems, 2014, pp. 172–178.
[56] M. Magnani, D. Montesi, G. Nunziante, and L. Rossi, “Conversation Retrieval from Twitter,” in Lecture Notes in Computer Science, vol. 6611 LNCS, 2011, pp. 780–783.
[57] P. Cogan, M. Andrews, M. Bradonjic, W. S. Kennedy, A. Sala, and G. Tucci, “Reconstruction and analysis of Twitter conversation graphs,” in Proceedings of the First ACM International Workshop on Hot Topics on Interdisciplinary Social Networks Research, HotSocial, 2012, pp. 25–31.
[58] S. Pace, S. Buzzanca, and L. Fratocchi, “The structure of conversations on social networks: Between dialogic and dialectic threads,” Int. J. Inf. Manage., vol. 36, no. 6, pp. 1144–1151, Dec. 2016.
[59] F. Yu, Q. Liu, S. Wu, L. Wang, and T. Tan, “Attention-based convolutional approach for misinformation identification from massive and noisy microblog posts,” Comput. Secur., vol. 83, pp. 106–121, Jun. 2019.
[60] J. Ma et al., “Detecting rumors from microblogs with recurrent neural networks,” in Proceedings of the Twenty-Fifth International Joint Conference on Artificial Intelligence, 2016, pp. 3818–3824.
[61] H. Guo, J. Cao, Y. Zhang, J. Guo, and J. Li, “Rumor Detection with Hierarchical Social Attention Network,” in Proceedings of the 27th ACM International Conference on Information and Knowledge Management, 2018, pp. 943–951.
[62] Z. Zhao, P. Resnick, and Q. Mei, “Enquiring Minds: Early Detection of Rumors in Social Media from Enquiry Posts,” in Proceedings of the 24th International Conference on World Wide Web, 2015, pp. 1395–1405.
[63] T. N. Nguyen, C. Li, and C. Niederée, “On Early-Stage Debunking Rumors on Twitter: Leveraging the Wisdom of Weak Learners,” in Lecture Notes in Computer Science, vol. 10540 LNCS, Cham: Springer International Publishing, 2017, pp. 141–158.
[64] W. L. Hamilton, R. Ying, and J. Leskovec, “Inductive representation learning on large graphs,” in Proceedings of the 31st Conference on Neural Information Processing Systems (NIPS 2017), 2017.
[65] A. Zubiaga, M. Liakata, R. Procter, G. Wong Sak Hoi, and P. Tolmie, “Analysing How People Orient to and Spread Rumours in Social Media by Looking at Conversational Threads,” PLoS One, vol. 11, no. 3, p. e0150989, Mar. 2016.
[66] J. Ma, W. Gao, and K.-F. Wong, “Rumor Detection on Twitter with Tree-structured Recursive Neural Networks,” in Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2018, vol. 1, pp. 1980–1989.
[67] J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding,” in Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 2019, pp. 4171–4186.
[68] P. Mishra, M. Del Tredici, H. Yannakoudakis, and E. Shutova, “Abusive Language Detection with Graph Convolutional Networks,” in Proceedings of the 2019 Conference of the North, 2019, no. 2018, pp. 2145–2150.
[69] L. M. Aiello, A. Barrat, R. Schifanella, C. Cattuto, B. Markines, and F. Menczer, “Friendship prediction and homophily in social media,” ACM Trans. Web, vol. 6, no. 2, 2012.
[70] M. McPherson, L. Smith-Lovin, and J. M. Cook, “Birds of a feather: Homophily in social networks,” Annu. Rev. Sociol., vol. 27, no. January, pp. 415–444, 2001.
[71] R. S. Solomon, S. Pykl, A. Das, B. Gamback, and T. Chakraborty, “Understanding the Psycho-Sociological Facets of Homophily in Social Network Communities,” IEEE Comput. Intell. Mag., vol. 14, no. 2, pp. 28–40, 2019.
[72] J. Li, S. Ni, and H. Y. Kao, “Birds of a Feather Rumor Together? Exploring Homogeneity and Conversation Structure in Social Media for Rumor Detection,” IEEE Access, vol. 8, pp. 212865–212875, 2020.
[73] N. K. Ahmed et al., “Inductive Representation Learning in Large Attributed Graphs,” arXiv Prepr., p. arXiv:1710.09471, 2017.
[74] A. Grover and J. Leskovec, “node2vec: Scalable Feature Learning for Networks,” in Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2016, pp. 855–864.
[75] T. Mikolov, K. Chen, G. Corrado, and J. Dean, “Efficient estimation of word representations in vector space,” in Proceedings of the 1st International Conference on Learning Representations, ICLR, Workshop Track Proceedings, 2013.
[76] Y. Liu and Y. F. B. Wu, “Early detection of fake news on social media through propagation path classification with recurrent and convolutional networks,” in Proceedings of the 32nd AAAI Conference on Artificial Intelligence,AAAI, 2018, no. December, pp. 354–361.
[77] K. Shu, L. Cui, S. Wang, D. Lee, and H. Liu, “Defend: Explainable fake news detection,” in Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2019, pp. 395–405.
[78] M. Gardner et al., “AllenNLP: A Deep Semantic Natural Language Processing Platform,” in Proceedings of Workshop for NLP Open Source Software, 2019, pp. 1–6.
[79] J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding,” in Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 2019, no. Mlm, pp. 4171–4186.
[80] J. Li, Y. Sujana, and H.-Y. Kao, “Exploiting Microblog Conversation Structures to Detect Rumors,” in Proceedings ofthe 28th International Conference on Computational Linguistics, 2021, pp. 5420–5429.