跳到主要內容

臺灣博碩士論文加值系統

(44.200.140.218) 您好!臺灣時間:2024/07/26 00:17
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果 :::

詳目顯示

我願授權國圖
: 
twitterline
研究生:陳泓諺
研究生(外文):CHEN, HONG-YAN
論文名稱:基於兩階段模型偵測新冠疫情下的中文冒犯性及仇恨言論
論文名稱(外文):A Two-Stage Model to Detect Abusive Language and Hate Speech in Chinese in the Context of the COVID-19 Pandemic
指導教授:王貞淑王貞淑引用關係
指導教授(外文):WANG, CHEN-SHU
口試委員:蕭文龍周棟祥王貞淑
口試委員(外文):SHIAU, WEN-LUNGCHOU, TUNG-HSIANGWANG, CHEN-SHU
口試日期:2023-05-31
學位類別:碩士
校院名稱:國立臺北科技大學
系所名稱:資訊與財金管理系
學門:商業及管理學門
學類:財務金融學類
論文種類:學術論文
論文出版年:2023
畢業學年度:111
語文別:中文
論文頁數:73
中文關鍵詞:冒犯性言論仇恨言論新冠肺炎機器學習自然語言處理生成式AI
外文關鍵詞:Abusive LanguageHate SpeechCOVID-19Machine LearningNatural Language ProcessingGenerative AI
相關次數:
  • 被引用被引用:0
  • 點閱點閱:120
  • 評分評分:
  • 下載下載:0
  • 收藏至我的研究室書目清單書目收藏:0
2020年的新冠疫情帶來了嚴重的健康和經濟危機,深刻地改變了人們的生活。並且因為疫情帶來的創傷和生活劇變,也在社群媒體平台上引發了大量的冒犯性和仇恨言論。這些具有傷害性的言論在網路上造成了族群對立、社會分化,甚至導致線下仇恨犯罪。因此,社群媒體上的冒犯性和仇恨言論已然成為當代社會日益嚴峻並且迫切需要解決的議題。然而,在這個眾聲喧嘩的時代,利用傳統的人工辨識方法來管理網路上的每分每秒產生的海量公共輿論資料變得不切實際。因此,利用機器學習技術來偵測線上的冒犯性和仇恨言論已是自然語言處理領域的一個日趨重要研究主題。本研究分別從人工標記建立中文冒犯性及仇恨言論資料集及仇恨言論字典、建立冒犯性及仇恨言論偵測模型及建立生成式AI仇恨言論改寫建議流程,在學術及實務面上做出貢獻。在字典方面,本研究擴充了中文不雅字字典及政治仇恨字典,新增了30個針對新冠疫情所產生的冒犯字及仇恨字。在冒犯性及仇恨言論偵測模型方面,本研究建立之兩階段模型,分別利用了支持向量機(SVM)、長短期記憶網路(LSTM)、雙向長短期記憶網路(Bi-LSTM)和BERT。經實驗比較後發現,BERT分類模型表現最佳,第一階段模型偵測冒犯性言論的準確率達到了94.42%,第二階段模型偵測仇恨言論的準確率達到了81.48%。最後運用生成式AI改寫仇恨言論,並利用LDA主題模型分析改寫前後差異,發現改寫後的言論能夠在保留原始主題及語意的情況下大幅減輕言論的仇恨程度,此發現可提供使用者在發送仇恨言論前提供可保留使用者語意的改寫建議以協助創造更友善和健康的公共輿論環境。
The outbreak of the COVID-19 pandemic in 2020 brought about severe health and economic crises, profoundly altering people's lives. Consequently, the trauma and upheaval caused by the pandemic led to a surge in abusive language and hate speech on social media platforms. These harmful expressions have resulted in societal divisions, intergroup conflicts, and even offline hate crimes. As a consequence, the propagation of abusive language and hate speech on social media has become an increasingly pressing and urgent issue in contemporary society.
In this era of overwhelming digital communication, employing traditional manual methods to manage the massive volume of public discourse data generated on the internet has become impractical. Therefore, utilizing machine learning techniques to detect abusive language and hate speech online has emerged as a critically important research topic in the field of natural language processing.
This study aims to make academic and practical contributions by creating datasets and dictionaries for abusive language and hate speech in the Chinese language through manual annotation. Also establishes models for detecting abusive language and hate speech and develops a process for generating AI-driven suggestions to rewrite hate speech.
Regarding the dictionaries, this research expands the existing Chinese profanity and political hate speech dictionaries by adding 30 terms specifically related to abusive language and hate speech arising from the COVID-19 pandemic.
In terms of detecting abusive language and hate speech, a two-stage model is proposed, incorporating Support Vector Machines (SVM), Long Short-Term Memory Networks (LSTM), Bidirectional Long Short-Term Memory Networks (Bi-LSTM), and BERT. Experimental comparisons reveal that the BERT classification model exhibits the best performance, achieving an accuracy of 94.42% in detecting abusive language in the first stage, and 81.48% accuracy in detecting hate speech in the second stage.
Finally, the study utilizes a generative AI model to rewrite hate speech and employs LDA topic modeling to analyze the differences between the original and rewritten content. The findings demonstrate that the rewritten speech significantly mitigates the hate level while preserving the essence and semantic content of the original text. This discovery provides users with suggestions for rewriting hateful expressions while retaining their intended meaning, thereby fostering a more friendly and healthy public discourse environment.

摘要 i
ABSTRACT iii
致謝 v
目錄 vi
表目錄 viii
圖目錄 x
第一章 緒論 1
1.1 研究背景與動機 1
1.2 研究目的 8
1.3 研究架構 9
第二章 文獻探討 11
2.1 仇恨言論(Hate Speech)及冒犯性言論(Abusive Language) 11
2.1.1 仇恨言論及冒犯性言論的定義 11
2.1.2 仇恨言論管制現況 14
2.2 文字探勘 16
2.3 機器學習 17
第三章 研究方法 19
3.1 資料蒐集 20
3.1.1 批踢踢實業坊(PTT) 20
3.1.2 資料篩選 21
3.2 引入字典 23
3.3 資料標註 25
3.4 模型建置 26
3.4.1 斷詞及去除停用詞 26
3.4.2 轉換為詞向量 27
3.4.3 SVM 28
3.4.4 長短期記憶模型(LSTM) 28
3.4.5 雙向長短期記憶模型(Bi-LSTM) 29
3.4.6 BERT 30
3.4.7 兩階段模型 31
3.4.8 模型評估 32
3.5 改寫及LDA 35
第四章 實驗設計與分析結果 37
4.1 資料描述與處理 37
4.2 實驗設計 38
4.2.1 實驗一:資料標註與一致性檢驗 39
4.2.2 實驗二:使用兩階段模型偵測冒犯性及仇恨言論 41
4.2.3 實驗三:仇恨言論改寫及LDA模型 60
第五章 結論與未來展望 66
5.1 結論 66
5.2 未來展望 67
參考文獻 69


1.Alonso, P., Saini, R., & Kovács, G. (2020). Hate speech detection using transformer ensembles on the hasoc dataset. Speech and Computer: 22nd International Conference, SPECOM 2020, St. Petersburg, Russia, October 7–9, 2020, Proceedings,
2.Anderson, K. F. (2013). Diagnosing discrimination: Stress from perceived racism and the mental and physical health effects. Sociological Inquiry, 83(1), 55-81.
3.Badjatiya, P., Gupta, S., Gupta, M., & Varma, V. (2017). Deep learning for hate speech detection in tweets. Proceedings of the 26th international conference on World Wide Web companion,
4.Bisht, A., Singh, A., Bhadauria, H., & Virmani, J. (2020). Detection of hate speech and offensive language in twitter data using lstm model. In Recent trends in image and signal processing in computer vision (pp. 243-264). Springer.
5.Boutyline, A., & Willer, R. (2017). The social structure of political echo chambers: Variation in ideological homophily in online networks. Political psychology, 38(3), 551-569.
6.Caselli, T., Basile, V., Mitrović, J., Kartoziya, I., & Granitzer, M. (2020). I feel offended, don’t be abusive! implicit/explicit messages in offensive and abusive language. Proceedings of the 12th language resources and evaluation conference,
7.Chetty, N., & Alathur, S. (2018). Hate speech review in the context of online social networks. Aggression and violent behavior, 40, 108-118.
8.Davidson, T., Warmsley, D., Macy, M., & Weber, I. (2017). Automated hate speech detection and the problem of offensive language. Proceedings of the international AAAI conference on web and social media,
9.Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2018). Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805.
10.Faisal, D. R., & Mahendra, R. (2022). Two-Stage Classifier for COVID-19 Misinformation Detection Using BERT: a Study on Indonesian Tweets. arXiv preprint arXiv:2206.15359.
11.Fan, L., Yu, H., & Yin, Z. (2020). Stigmatization in social media: Documenting and analyzing hate speech for COVID‐19 on Twitter. Proceedings of the Association for Information Science and Technology, 57(1), e313.
12.Fernández, A., López, V., Galar, M., Del Jesus, M. J., & Herrera, F. (2013). Analysing the classification of imbalanced data-sets with multiple classes: Binarization techniques and ad-hoc approaches. Knowledge-based systems, 42, 97-110.
13.Fleiss, J. L. (1971). Measuring nominal scale agreement among many raters. Psychological bulletin, 76(5), 378.
14.Fortuna, P., & Nunes, S. (2018). A survey on automatic detection of hate speech in text. ACM Computing Surveys (CSUR), 51(4), 1-30.
15.Graves, A., & Schmidhuber, J. (2005). Framewise phoneme classification with bidirectional LSTM and other neural network architectures. Neural networks, 18(5-6), 602-610.
16.Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural computation, 9(8), 1735-1780.
17.Khan, S., Fazil, M., Sejwal, V. K., Alshara, M. A., Alotaibi, R. M., Kamal, A., & Baig, A. R. (2022). BiCHAT: BiLSTM with deep CNN and hierarchical attention for hate speech detection. Journal of King Saud University-Computer and Information Sciences, 34(7), 4335-4344.
18.Langton, R. (1993). Speech acts and unspeakable acts. Philosophy & public affairs, 293-330.
19.Liu, H., Burnap, P., Alorainy, W., & Williams, M. (2020). Scmhl5 at TRAC-2 shared task on aggression identification: bert based ensemble learning approach.
20.Maitra, I., & McGowan, M. K. (2012). Speech and harm: Controversies over free speech. Oxford University Press.
21.Matsuda, M. J. (1988). Public response to racist speech: Considering the victim's story. Mich. L. Rev., 87, 2320.
22.Mozafari, M., Farahbakhsh, R., & Crespi, N. (2019). A BERT-based transfer learning approach for hate speech detection in online social media. International Conference on Complex Networks and Their Applications,
23.Nichols, T. R., Wisner, P. M., Cripe, G., & Gulabchand, L. (2010). Putting the kappa statistic to use. The Quality Assurance Journal, 13(3-4), 57-61.
24.Niemann, M., Riehle, D. M., Brunk, J., & Becker, J. (2019). What is abusive language? Multidisciplinary International Symposium on Disinformation in Open Online Media,
25.Niemann, M., Riehle, D. M., Brunk, J., & Becker, J. (2020). What is abusive language? Multidisciplinary International Symposium on Disinformation in Open Online Media,
26.Nikhil, N., Pahwa, R., Nirala, M. K., & Khilnani, R. (2018). Lstms with attention for aggression detection. Proceedings of the first workshop on trolling, aggression and cyberbullying (TRAC-2018),
27.Nobata, C., Tetreault, J., Thomas, A., Mehdad, Y., & Chang, Y. (2016). Abusive language detection in online user content. Proceedings of the 25th international conference on world wide web,
28.Saha, P., Das, M., Mathew, B., & Mukherjee, A. (2023). Hate Speech: Detection, Mitigation and Beyond. Proceedings of the Sixteenth ACM International Conference on Web Search and Data Mining,
29.Sontag, S. (1997). AIDS and its metaphors. The disability studies reader, 232-240.
30.Sontag, S., & Broun, H. H. (1977). Illness as metaphor. Farrar, Straus New York.
31.Subyantoro, S., & Apriyanto, S. (2020). Impoliteness in Indonesian language hate speech on social media contained in the Instagram account. Journal of Advances in Linguistics, 11(2), 36-46.
32.Swamy, S. D., Jamatia, A., & Gambäck, B. (2019). Studying generalisability across abusive language detection datasets. Proceedings of the 23rd conference on computational natural language learning (CoNLL),
33.Tan, W., Yao, Q., & Liu, J. (2022). Two-Stage COVID19 Classification Using BERT Features. arXiv preprint arXiv:2206.14861.
34.Tiţa, T., & Zubiaga, A. (2021). Cross-lingual Hate Speech Detection using Transformer Models. arXiv preprint arXiv:2111.00981.
35.Vidgen, B., & Derczynski, L. (2020). Directions in abusive language training data, a systematic review: Garbage in, garbage out. Plos one, 15(12), e0243300.
36.Vishwamitra, N., Hu, R. R., Luo, F., Cheng, L., Costello, M., & Yang, Y. (2020). On analyzing covid-19-related hate speech using bert attention. 2020 19th IEEE International Conference on Machine Learning and Applications (ICMLA),
37.Wang, C.-C., Day, M.-Y., & Wu, C.-L. (2022). Political Hate Speech Detection and Lexicon Building: A Study in Taiwan. IEEE Access, 10, 44337-44346.
38.Waseem, Z., & Hovy, D. (2016). Hateful symbols or hateful people? predictive features for hate speech detection on twitter. Proceedings of the NAACL student research workshop,
39.Weber, A. (2009). Manual on hate speech. Council of Europe.
40.Yang, H., & Lin, C.-J. (2020). Tocp: A dataset for chinese profanity processing. Proceedings of the Second Workshop on Trolling, Aggression and Cyberbullying,
41.Zanini, N., & Dhawan, V. (2015). Text Mining: An introduction to theory and some applications. Research Matters, 19, 38-45.
42.Zhang, Z., Robinson, D., & Tepper, J. (2018). Detecting hate speech on twitter using a convolution-gru based deep neural network. European semantic web conference,
43.廖福特. (2015). 什麼是仇恨言論, 應否及如何管制: 歐洲人權法院相關判決分析. EurAmerica, 45(4).


電子全文 電子全文(網際網路公開日期:20280717)
QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top
無相關期刊