Skip navigation

DSpace

機構典藏 DSpace 系統致力於保存各式數位資料(如:文字、圖片、PDF)並使其易於取用。

點此認識 DSpace
DSpace logo
English
中文
  • 瀏覽論文
    • 校院系所
    • 出版年
    • 作者
    • 標題
    • 關鍵字
  • 搜尋 TDR
  • 授權 Q&A
    • 我的頁面
    • 接受 E-mail 通知
    • 編輯個人資料
  1. NTU Theses and Dissertations Repository
  2. 社會科學院
  3. 新聞研究所
請用此 Handle URI 來引用此文件: http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/86111
完整後設資料紀錄
DC 欄位值語言
dc.contributor.advisor謝吉隆(Ji-Lung Hsieh)
dc.contributor.authorTzu-Hsuan Tsengen
dc.contributor.author曾子軒zh_TW
dc.date.accessioned2023-03-19T23:37:22Z-
dc.date.copyright2022-09-12
dc.date.issued2022
dc.date.submitted2022-09-08
dc.identifier.urihttp://tdr.lib.ntu.edu.tw/jspui/handle/123456789/86111-
dc.description.abstract因為能夠反應社會上的價值觀,再加上其性別化的訊息,還有新聞性的本質,流行雜誌時常是性別研究的研究對象。先前的研究表明,詞嵌入模型能夠揭示媒體中性別之間的微妙偏見。本研究以針對特定性別的時尚雜誌為研究對象,不僅呈現出不同文本中各自的性別差異,更透過比較雜誌之間的偏見,對性別研究做出貢獻。 蒐集女性雜誌和男性雜誌超過 10 年的雜誌文章後,本研究先以主題模型查看時尚雜誌中的主題分佈,還有時間變化與性別差異。並分別構建了詞嵌入模型,接著比較了它們在性別偏見方面的差異。除了用於檢查性別偏見的詞彙表,如職業和形容詞,還考慮了出現在女權主義辯論中的其他話題,如身體部位、浪漫愛情、性愛、快樂等。 研究結果發現,個別雜誌中出現了性別偏見,這並不是令人意外之事。然而,進一步發現,在針對性別的雜誌之間,表現出的性別偏見有著不同方向,例如外表有關的形容詞,以及身體部位,在女性和男性之間,表現出明顯差異。zh_TW
dc.description.abstractAs a subject of a gender study, popular magazines are distinctively scrutinized due to their journalistic essences reflecting the social value and often sexualized messages. Previous research shows that word embedding models can reveal subtle bias between gender in media. This research contributes to gender studies by not only presenting gender differences within texts but also comparing differences of bias between gender-targeted magazines. With 10-year-long magazine articles from a woman’s magazine and a man’s magazine, I construct word embeddings models respectively and compare their divergence of gender bias. In addition to word lists used to inspect gender bias, such as vocations and adjectives, other topics appearing in the feminism debate have been considered, such as body parts, romantic love, sexiness, and so on. I found gender bias appears in the individual magazines, which is not a surprise. However, it is also found that gender biases between magazines are different. Appearance-related adjectives and body parts show remarkably different directions between males and females.en
dc.description.provenanceMade available in DSpace on 2023-03-19T23:37:22Z (GMT). No. of bitstreams: 1
U0001-0609202201032700.pdf: 2437924 bytes, checksum: fef3e3592d307cbe99be38bdb6d39456 (MD5)
Previous issue date: 2022
en
dc.description.tableofcontentsTable of Contents Introduction 1 Literature Review 3 Gender Bias in Magazines 5 Word Embeddings and Bias 9 Research Questions 12 Data and Methods 14 Data Collection 14 Content Analysis with Topic Modeling 15 Exploring Gender Bias with Word Embeddings 17 Comparing Gender Bias between Word Embeddings Models 23 Checking Quality of Word Embeddings Models 25 Results 29 RQ1: Themes of Content 29 RQ2: Within-Bias 32 RQ3: Between-Bias 35 Discussion 40 Topics and Gender 40 Bias and Gender 41 Contributions 44 References 46 Appendix 60 appendix 1: word lists of gender vectors 60 appendix 2: topic model 63 appendix 3: word lists of WEAT test 66 appendix 4: Complete result of direction difference 66 Table of Contents - Figure Figure 1. 30 topics and corresponding words with the highest prevalence within topics. 16 Figure 2. Measuring gender bias with distance between words and gender vectors. 18 Figure 3. Measuring implicit bias with IAT. 20 Figure 4. Examining bias in word embeddings with WEAT. 22 Figure 5. Examining bias in word embeddings with ECT. 22 Figure 6. Examining bias between word embeddings with paired McNemar’s Chi-square test. 24 Figure 7. Examining bias between word embeddings with cross-corpus ECT. 25 Figure 8. The share of topics across years. 30 Figure 9. Difference of point estimates for 30 topics based on Bayesian regression models. 31 Table of Contents - Table Table 1. Results of similarity test. 28 Table 2. Results of analogy test. 29 Table 3. Themes of topics. 30 Table 4. Top 10 words closest to “he vectors” and “she vectors” in magazines. 33 Table 5. Summary of the result of Pearson's correlation between human-labeled gender-skewness and word embeddings models 34 Table 6. WEAT results. 35 Table 7. ECT results. 35 Table 8. Summary of the result of paired McNemar’s Chi-square test with adjectives as test set. 36 Table 9. Summary of the result of paired McNemar’s Chi-square test with professions lists as test set. 36 Table 10. cross-corpus-ECT results. 36 Table 11. Leaning of thematic word lists between magazines. 37 Table 12. Co-occurrence words of sexy(性感) and body shape(身材). 39 Table 13. Co-occurrence words of relationships(感情) and happiness(快樂). 40
dc.language.isoen
dc.title她的偏見不是他的偏見:比較女性與男性雜誌的性別偏見zh_TW
dc.titleHer bias is not his bias: A comparison of embedded gender bias between gender-targeted magazinesen
dc.typeThesis
dc.date.schoolyear110-2
dc.description.degree碩士
dc.contributor.author-orcid0000-0002-5406-1314
dc.contributor.coadvisor劉好迪(Adrian Rauchfleisch)
dc.contributor.oralexamcommittee陳正賢(Cheng-Hsien Chen),鄧志松(Chih-sung Teng)
dc.subject.keyword性別偏見,時尚雜誌,詞嵌入,自然語言處理,zh_TW
dc.subject.keywordGender bias,Fashion Magazine,Word Embeddings,Natural Language Processing,en
dc.relation.page67
dc.identifier.doi10.6342/NTU202203184
dc.rights.note同意授權(全球公開)
dc.date.accepted2022-09-08
dc.contributor.author-college社會科學院zh_TW
dc.contributor.author-dept新聞研究所zh_TW
dc.date.embargo-lift2022-09-12-
顯示於系所單位:新聞研究所

文件中的檔案:
檔案 大小格式 
U0001-0609202201032700.pdf2.38 MBAdobe PDF檢視/開啟
顯示文件簡單紀錄


系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。

社群連結
聯絡資訊
10617臺北市大安區羅斯福路四段1號
No.1 Sec.4, Roosevelt Rd., Taipei, Taiwan, R.O.C. 106
Tel: (02)33662353
Email: ntuetds@ntu.edu.tw
意見箱
相關連結
館藏目錄
國內圖書館整合查詢 MetaCat
臺大學術典藏 NTU Scholars
臺大圖書館數位典藏館
本站聲明
© NTU Library All Rights Reserved