請用此 Handle URI 來引用此文件：
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/73032完整後設資料紀錄
| DC 欄位 | 值 | 語言 |
|---|---|---|
| dc.contributor.advisor | 李琳山(Lin-shan Lee) | |
| dc.contributor.author | Chia-Hsuan Lee | en |
| dc.contributor.author | 李佳軒 | zh_TW |
| dc.date.accessioned | 2021-06-17T07:14:38Z | - |
| dc.date.available | 2019-07-17 | |
| dc.date.copyright | 2019-07-17 | |
| dc.date.issued | 2019 | |
| dc.date.submitted | 2019-07-16 | |
| dc.identifier.citation | [1] Pranav Rajpurkar, Jian Zhang, Konstantin Lopyrev, and Percy Liang, “Squad: 100,000+ questions for machine comprehension of text,” in Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, 2016, pp. 2383–2392.
[2] Chia-Hsuan Lee, Szu-Lin Wu, Chi-Liang Liu, and Hung-yi Lee, “Spo- ken squad: A study of mitigating the impact of speech recognition errors on listening comprehension,” Proc. Interspeech 2018, pp. 3459–3463, 2018. [3] Chia-Hsuan Lee, Shang-Ming Wang, Huan-Cheng Chang, and Hung-Yi Lee, “Odsqa: Open-domain spoken question answering dataset,” in 2018 IEEE Spoken Language Technology Workshop (SLT). IEEE, 2018, pp. 949–956. [4] Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin, “Attention is all you need,” in Advances in neural information processing systems, 2017, pp. 5998–6008. [5] Tri Nguyen, Mir Rosenberg, Xia Song, Jianfeng Gao, Saurabh Ti- wary, Rangan Majumder, and Li Deng, “Ms marco: A human generated machine reading comprehension dataset,” arXiv preprint arXiv:1611.09268, 2016. [6] AdamsWeiYu,DavidDohan,Minh-ThangLuong,RuiZhao,KaiChen, Mohammad Norouzi, and Quoc V Le, “Qanet: Combining local con- volution with global self-attention for reading comprehension,” arXiv preprint arXiv:1804.09541, 2018. [7] Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio, “Generative adversarial nets,” in Advances in neural information pro- cessing systems, 2014, pp. 2672–2680. [8] Bo-Hsiang Tseng, Sheng-syun Shen, Hung-Yi Lee, and Lin-Shan Lee, “Towards machine comprehension of spoken content: Initial toefl listen- ing comprehension test by machine,” Interspeech 2016, pp. 2731–2735, 2016. [9] Paul Lamere, Philip Kwok, Evandro Gouvea, Bhiksha Raj, Rita Singh, and William Walker, “The cmu sphinx-4 speech recognition system,” . [10] Chih Chieh Shao, Trois Liu, Yuting Lai, Yiying Tseng, and Sam Tsai, “Drcd: a chinese machine reading comprehension dataset,” arXiv preprint arXiv:1806.00920, 2018. [11] Thang Luong, Richard Socher, and Christopher Manning, “Better word representations with recursive neural networks for morphology,” in Pro- ceedings of the Seventeenth Conference on Computational Natural Lan- guage Learning, 2013, pp. 104–113. [12] Piotr Bojanowski, Edouard Grave, Armand Joulin, and Tomas Mikolov, “Enriching word vectors with subword information,” Transactions of the Association for Computational Linguistics, vol. 5, pp. 135–146, 2017. [13] Minjoon Seo, Aniruddha Kembhavi, Ali Farhadi, and Hannaneh Ha- jishirzi, “Bidirectional attention flow for machine comprehension,” arXiv preprint arXiv:1611.01603, 2016. [14] Wenhui Wang, Nan Yang, Furu Wei, Baobao Chang, and Ming Zhou, “Gated self-matching networks for reading comprehension and question answering,” in Proceedings of the 55th Annual Meeting of the Associa- tion for Computational Linguistics (Volume 1: Long Papers), 2017, pp. 189–198. [15] Hsin-Yuan Huang, Chenguang Zhu, Yelong Shen, and Weizhu Chen, “Fusionnet: Fusing via fully-aware attention with application to machine comprehension,” arXiv preprint arXiv:1711.07341, 2017. [16] Danqi Chen, Adam Fisch, Jason Weston, and Antoine Bordes, “Read- ing wikipedia to answer open-domain questions,” in Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2017, pp. 1870–1879. [17] Minghao Hu, Yuxing Peng, Zhen Huang, Xipeng Qiu, Furu Wei, and Ming Zhou, “Reinforced mnemonic reader for machine reading com- prehension,” in Proceedings of the 27th International Joint Conference on Artificial Intelligence. AAAI Press, 2018, pp. 4099–4106. [18] Yaroslav Ganin, Evgeniya Ustinova, Hana Ajakan, Pascal Germain, Hugo Larochelle, Franc ̧ois Laviolette, Mario Marchand, and Victor Lempitsky, “Domain-adversarial training of neural networks,” The Jour- nal of Machine Learning Research, vol. 17, no. 1, pp. 2096–2030, 2016. [19] Yusuke Shinohara, “Adversarial multi-task learning of deep neural net- works for robust speech recognition.,” in INTERSPEECH, 2016, pp. 2369–2372. [20] Zhilin Yang, Ruslan Salakhutdinov, and William W Cohen, “Trans- fer learning for sequence tagging with hierarchical recurrent networks,” arXiv preprint arXiv:1703.06345, 2017. [21] David McClosky, Eugene Charniak, and Mark Johnson, “Automatic do- main adaptation for parsing,” in Human Language Technologies: The 2010 Annual Conference of the North American Chapter of the Asso- ciation for Computational Linguistics. Association for Computational Linguistics, 2010, pp. 28–36. [22] Laura Chiticariu, Rajasekar Krishnamurthy, Yunyao Li, Frederick Reiss, and Shivakumar Vaithyanathan, “Domain adaptation of rule-based an- notators for named-entity recognition tasks,” in Proceedings of the 2010 conference on empirical methods in natural language processing. Asso- ciation for Computational Linguistics, 2010, pp. 1002–1012. [23] Rudolf Kadlec, Ondrej Bajgar, and Jan Kleindienst, “From particular to general: A preliminary case study of transfer learning in reading com- prehension,” in Machine Intelligence Workshop, NIPS, 2016. [24] Sewon Min, Minjoon Seo, and Hannaneh Hajishirzi, “Question answer- ing through transfer learning from large fine-grained supervision data,” arXiv preprint arXiv:1702.02171, 2017. [25] Georg Wiese, Dirk Weissenborn, and Mariana Neves, “Neural do-main adaptation for biomedical question answering,” arXiv preprint arXiv:1706.03610, 2017. [26] David Golub, Po-Sen Huang, Xiaodong He, and Li Deng, “Two-stage synthesis networks for transfer learning in machine comprehension,” arXiv preprint arXiv:1706.09789, 2017. [27] Bernardo Magnini, Danilo Giampiccolo, Pamela Forner, Christelle Ay- ache, Valentin Jijkoun, Petya Osenova, Anselmo Pen ̃as, Paulo Rocha, Bogdan Sacaleanu, and Richard Sutcliffe, “Overview of the clef 2006 multilingual question answering track,” in Workshop of the Cross- Language Evaluation Forum for European Languages. Springer, 2006, pp. 223–256. [28] Yutaka Sasaki, Hsin-Hsi Chen, Kuang-hua Chen, and Chuan-Jie Lin, “Overview of the ntcir-5 cross-lingual question answering task (clqa1).,” in NTCIR, 2005. [29] Tatsunori Mori and Kousuke Takahashi, “A method of cross-lingual question-answering based on machine translation and noun phrase trans- lation using web documents,” in Proceedings of NTCIR-6 Workshop, Tokyo, Japan, 2007. [30] Johan Bos and Malvina Nissim, “Cross-lingual question answering by answer translation.,” in CLEF (Working Notes), 2006. [31] Saif M Mohammad, Mohammad Salameh, and Svetlana Kiritchenko, “How translation alters sentiment,” Journal of Artificial Intelligence Re- search, vol. 55, pp. 95–130, 2016. [32] Evgeny A Stepanov, Ilya Kashkarev, Ali Orkan Bayer, Giuseppe Ric- cardi, and Arindam Ghosh, “Language style and domain adaptation for cross-language slu porting,” in Automatic Speech Recognition and Understanding (ASRU), 2013 IEEE Workshop on. IEEE, 2013, pp. 144– 149. [33] FernandoGarc ́ıa,Llu ́ısFHurtado,EncarnaSegarra,EmilioSanchis,and Giuseppe Riccardi, “Combining multiple translation systems for spoken language understanding portability,” in Spoken Language Technology Workshop (SLT), 2012 IEEE. IEEE, 2012, pp. 194–198. [34] Xiaodong He, Li Deng, Dilek Hakkani-Tur, and Gokhan Tur, “Multi- style adaptive training for robust cross-lingual spoken language under- standing,” in Acoustics, Speech and Signal Processing (ICASSP), 2013 IEEE International Conference on. IEEE, 2013, pp. 8342–8346. [35] XilunChen,YuSun,BenAthiwaratkun,ClaireCardie,andKilianWein- berger, “Adversarial deep averaging networks for cross-lingual senti- ment classification,” arXiv preprint arXiv:1606.01614, 2016. [36] Joo-Kyung Kim, Young-Bum Kim, Ruhi Sarikaya, and Eric Fosler- Lussier, “Cross-lingual transfer learning for pos tagging without cross- lingual resources,” in Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, 2017, pp. 2832–2838. [37] Alexis Conneau, Guillaume Lample, Marc’Aurelio Ranzato, Ludovic Denoyer,andHerve ́Je ́gou, “Wordtranslationwithoutparalleldata,” arXiv preprint arXiv:1710.04087, 2017. [38] Adam Trischler, Tong Wang, Xingdi Yuan, Justin Harris, Alessandro Sordoni, Philip Bachman, and Kaheer Suleman, “Newsqa: A machine comprehension dataset,” in Proceedings of the 2nd Workshop on Repre- sentation Learning for NLP, 2017, pp. 191–200. [39] Ishaan Gulrajani, Faruk Ahmed, Martin Arjovsky, Vincent Dumoulin, and Aaron C Courville, “Improved training of wasserstein gans,” in Advances in Neural Information Processing Systems, 2017, pp. 5767– 5777. [40] Guillaume Lample, Neil Zeghidour, Nicolas Usunier, Antoine Bordes, Ludovic Denoyer, et al., “Fader networks: Manipulating images by slid- ing attributes,” in Advances in Neural Information Processing Systems, 2017, pp. 5967–5976. | |
| dc.identifier.uri | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/73032 | - |
| dc.description.abstract | 本論文研究主題涵蓋兩大前瞻性方向:語音問答系統(Spoken Question Answering)和跨語言問答系統(Cross Lingual Question Answering)，實驗語料為英文以及中文，包含語音訊號以及語音辨識轉寫(Transcription)。
語音問答以英文維基百科的合成聲音以及中文維基百科的真實人聲為應用領域， 本論文蒐集建構了兩個大型問答集，提供了研究者們夠大量的高品質的資源，嘗試解決過去無法訓練複雜深層學習(Deep Learning)模型的問題。由於含有眾多語 音辨識錯誤，在標竿資料集上表現頂尖的問答系統在語音環境下表現，相較於純 文字環境下皆呈現大幅度退步。本論文使用聲學上的次詞單位去呈現文章中的詞，藉由次詞單位可以將辨識錯誤的文字與正確轉寫的文字之間的語意做連結， 讓問答系統可以得到較為貼近正確轉寫文字的資訊，實驗結果顯示不論在中文或 英文上，次詞單位都能夠有效提升模型的表現。 由於訓練深層學習的問答模型需要大量人工標注的資料，並且當前幾乎所有大型 問答資料集都是英文的，因此問答模型的發展在英文以外的語言上相當緩慢。本論文探討如何將英文資料裡的知識遷移到中文的問答模型上，亦即跨語言遷移學 習(Cross Lingual Transfer Learning)，首先，使用機器翻譯系統將英文資料集翻譯 成中文，作為額外的訓練資料，可以成功提升中文問答模型的表現。然而，不是所有語言之間都有高品質的機器翻譯系統。本論文接著提出一個只需要詞對詞雙 語詞典作為語言資源的模型，此模型引入了生成對抗學習，透過句子編碼器與語 言鑑別器之間的對抗，句子編碼器可以將不同語言文句的表徵(Representation)投 射到共享的高維向量空間上，因此可以讓模型同時從不同語言的資料中有效率的學習。 | zh_TW |
| dc.description.provenance | Made available in DSpace on 2021-06-17T07:14:38Z (GMT). No. of bitstreams: 1 ntu-108-R06944037-1.pdf: 5979510 bytes, checksum: ccc0b9150a2745b1c270a35dfb6f543e (MD5) Previous issue date: 2019 | en |
| dc.description.tableofcontents | i 中文摘要.......................................
ii 一、導論....................................... 1 1.1 背景研究.................................. 1 1.2 論文研究方向及貢獻 ........................... 2 1.3 章節安排.................................. 4 二、背景知識 .................................... 5 2.1 深層學習基礎觀念............................. 5 2.1.1 多層感知器 ............................ 5 2.1.2 卷積類神經網路.......................... 7 2.1.3 自專注神經網路(Self Attention Neural Network) . . . . . . . . 8 2.2 問答系統.................................. 10 2.2.1 背景介紹:文字問答系統 .................... 10 2.2.2 檢索為基礎之文字問答系統(非結構化文字集合中的網路文章) 14 2.2.3 知識為基礎之文字問答系統(結構化的知識庫) . . . . . . . . . 14 2.2.4 深度學習時代的文字問答系統:以問答網路(QANet)為例 . . . 15 2.2.5 語音問答系統 ........................... 16 2.3 生成對抗網路 ............................... 17 2.4 本章總結.................................. 19 三、語音問答系統.................................. 21 3.1 背景介紹.................................. 21 3.2 語料庫之蒐集 ............................... 22 3.2.1 語音合成系統:Spoken-SQuAD ................. 22 3.2.2 人工蒐集:ODSQA......................... 23 3.3 使用次詞單位緩解語音辨識錯誤之影響................. 26 3.3.1 詞典外問題與次詞單位...................... 26 3.3.2 次詞單位種類與簡述 ....................... 27 3.3.3 使用次詞單位緩解語音辨識錯誤 ................ 29 3.3.4 次詞單位的模型.......................... 29 3.4 實驗設計與結果 .............................. 31 3.4.1 英語的語音問答模型實驗 .................... 31 3.4.2 中文的語音問答模型實驗 .................... 34 3.5 本章總結.................................. 38 四、跨語言問答系統 ................................ 44 4.1 背景..................................... 44 4.1.1 跨語言問答系統.......................... 44 4.1.2 問答系統的遷移學習 ....................... 44 4.1.3 多語言問答系統.......................... 45 4.1.4 跨語言遷移學習.......................... 45 4.1.5 跨語言遷移學習的任務介紹 ................... 46 4.1.6 機器翻譯系統用於跨語言遷移學習 ............... 46 4.2 生成對抗網路用於跨語言遷移學習 ................... 47 4.2.1 雙語向量.............................. 47 4.2.2 語言相關(Language Dependent)模組與語言不相關(Language- Independent)模組 ......................... 48 4.2.3 對抗學習.............................. 49 4.3 實驗結果.................................. 50 4.3.1 資料集 ............................... 50 4.3.2 實驗設置.............................. 51 4.3.3 模型設計.............................. 52 4.3.4 參數選擇.............................. 53 4.3.5 基於機器翻譯的跨語言遷移學習 ................ 53 4.3.6 基於生成對抗網路的跨語言遷移學習.............. 54 4.3.7 結合機器翻譯方法與生成對抗網路方法 ............ 57 五、結論與展望 ................................... 61 5.1 總結..................................... 61 5.2 未來展望.................................. 62 參考文獻....................................... 63 | |
| dc.language.iso | zh-TW | |
| dc.subject | 語音問答 | zh_TW |
| dc.subject | 次詞 | zh_TW |
| dc.subject | 文字問答 | zh_TW |
| dc.subject | 遷移學習 | zh_TW |
| dc.subject | Text Question Answering | en |
| dc.subject | Speech Question Answering | en |
| dc.subject | Subword | en |
| dc.subject | Transfer Learning | en |
| dc.title | 使用次詞單位及遷移學習之跨語言及語音問答系統 | zh_TW |
| dc.title | Cross Lingual and Spoken Question Answering with Subword Units and Transfer Learning | en |
| dc.type | Thesis | |
| dc.date.schoolyear | 107-2 | |
| dc.description.degree | 碩士 | |
| dc.contributor.oralexamcommittee | 李宏毅(Hung-Yi Lee),陳信宏(Sin-Horng Chen),王小川,鄭秋豫 | |
| dc.subject.keyword | 文字問答,語音問答,次詞,遷移學習, | zh_TW |
| dc.subject.keyword | Text Question Answering,Speech Question Answering,Subword,Transfer Learning, | en |
| dc.relation.page | 69 | |
| dc.identifier.doi | 10.6342/NTU201901515 | |
| dc.rights.note | 有償授權 | |
| dc.date.accepted | 2019-07-16 | |
| dc.contributor.author-college | 電機資訊學院 | zh_TW |
| dc.contributor.author-dept | 資訊網路與多媒體研究所 | zh_TW |
| 顯示於系所單位： | 資訊網路與多媒體研究所 | |
文件中的檔案：
| 檔案 | 大小 | 格式 | |
|---|---|---|---|
| ntu-108-1.pdf 未授權公開取用 | 5.84 MB | Adobe PDF |
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。