簡易檢索 / 詳目顯示

回結果列表
研究生： 王敬嘉
Wang, Ching-Chia
論文名稱： 使用歌詞情緒與情境元素進行社群文章音樂推薦
Using Emotion and Scene Features of Lyrics for Social Article Music Recommendation
指導教授： 盧文祥
Lu, Wen-Hsiang
學位類別： 碩士
Master
系所名稱： 電機資訊學院 - 資訊工程學系
Department of Computer Science and Information Engineering
論文出版年： 2017
畢業學年度： 105
語文別： 英文
論文頁數： 60
中文關鍵詞： 音樂推薦 、歌詞分析 、音樂情緒 、音樂情境 、情境分類 、標記式隱含狄立克雷模型 、概念模型 、歌詞
外文關鍵詞： Music Recommendation, Lyrics Analysis, Music Emotion, Music Scene, Scene Classification, Labeled-Latent Dirichlet Allocation, Bag of Concepts, Lyrics
相關次數： 點閱：127  下載：1 
分享至:
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報

    • 隨著數位音樂市場的成長，發展自動化音樂推薦系統成為熱門主題，雖然許多串流公司利用音樂的音訊資料、後設數據與協同式過濾等技術，開發出符合使用者音樂喜好的音樂推薦系統，但以上技術無法分析並詮釋歌詞心境。然而，當使用者在社群網站分享音樂的目的是欲輔助理解生活相關情境文章時，使用者會傾向選擇歌詞描述心境與文章相關的歌曲，而非單純選擇符合其音樂喜好的歌曲。
    為了解決上述問題，我們提出基於歌詞情緒與情境的社群文章音樂推薦系統，針對一篇社群文章，此系統將分析歌詞心境後推薦相關歌曲清單。這篇論文中，我們將歌詞心境視為由歌詞情緒與歌詞情境組合而成，其中歌詞情緒代表一首歌曲心理上的主觀描述，而歌詞情境代表一首歌曲的客觀描述。我們利用中文廣義知網（Extended-HowNet）作為中文知識庫來抽取歌詞情緒與情境特徵。透過應用標記式隱含狄立克雷分布（Labeled-Latent Dirichlet Allocation），我們將情緒特徵表達為48種情緒主題的主題機率分布，而透過應用Bag-of-Concepts的概念模型以及調整自TF-IDF的CF-IDF方法，我們將情境特徵表達為概念階層的特徵向量。我們透過實驗證明同時考慮歌詞情緒與情境比起只使用其中一項特徵更能在針對社群文章的歌曲推薦上得到良好效果，最後我們的推薦系統在系統效能與系統偏好上都贏過使用word2vec的基準方法。

    With the potential of digital music industry, developing automatic music recommendation engines becomes a popular issue. In spite that audio information, music metadata and collaborative filtering offer recommendation results which match users’ music taste, none of them can cover the lyrical theme of a song. However, in the scenario of sharing related songs with user-generated articles about daily life on online social platforms, users tend to choose songs considering their lyrical theme.
    To solve the above problem, we present an Emotion-Scene-based Song Recommendation System which can recommend list of songs to an input article by analyzing lyrical theme. We consider lyrical theme as a combination of Emotion and Scene, the subjective and objective perspective of lyrics. By utilizing Extended-HowNet as knowledge base, we extract emotion and scene features of lyrics. Emotion feature is represented as probability of emotion topic distribution over 48 emotion topics using supervised topic modeling method, Labeled-Latent Dirichlet Allocation. Scene feature is represented as concept level representation using the idea of Bag-of-Concepts and a TF-IDF-liked method, CF-IDF.
    We show that using both emotion and scene features provide better recommendation results than merely consider one of the features. In the end our system outperforms a novel W2V baseline in both experiments of system performance and system preference.

    摘要 III Abstract V 致謝 VII Table of Contents VIII List of Tables X List of Figures XI Chapter 1 Introduction 1 1.1 Background 1 1.2 Motivation 2 1.3 Method 4 1.4 Contribution 5 1.5 Organization of this Dissertation 6 Chapter 2 Related Work 7 2.1 Studies on Exploring Song Lyrics 7 2.2 Studies on Music Emotion Discovery based on Lyrics 8 2.3 Studies on Music Topic Detection based on Lyrics 8 2.4 Studies on Music Recommendation using Lyrics 8 Chapter 3 Method 10 3.1 System Framework 10 3.2 Preliminaries 12 3.2.1 Extended-HowNet 12 3.2.2 Datasets and Preprocessing Steps 13 3.3 Emotion Feature Generation 15 3.3.1 Emotion Term Extracting 16 3.3.2 Emotion Topic Modeling 20 3.3.3 Emotion Feature Generating 25 3.4 Scene Features Generation 25 3.4.1 Scene Term Extracting 26 3.4.2 Scene Feature Generating 27 3.5 Scene Classification 30 3.6 Emotion-Scene based Song Recommendation 32 Chapter 4 Experiments 34 4.1 Dataset 34 4.2 Experiment of Scene Classification 35 4.2.1 Dataset for Scene Classification 35 4.2.2 Evaluation Metrics 37 4.2.3 Experiment Result 37 4.3 Experiment of Recommendation Parameter λ 38 4.3.1 Dataset for Evaluate Recommendation Parameter λ 38 4.3.2 Evaluation Metrics 38 4.3.3 Experiment Result 40 4.4 Evaluation of Emotion-Scene-based Song Recommendation 42 4.4.1 Dataset for Evaluating Emotion-Scene based Song Recommendation 42 4.4.2 Baseline 43 4.4.3 Evaluation Metrics 43 4.4.4 Experiment Result 48 Chapter 5 Conclusions 56 Reference 57

    [1] ‘Songwriting Tip: Lyrical Themes’, StudioPros Inc (2010), https://studiopros.com/songwriting-tip-lyrical-themes/
    [2] Lee, J. H., & Downie, J. S. (2004). Survey of music information needs, uses, and seeking behaviours: Preliminary findings. In Proceedings of the 5th international Symposium on Music Information Retrieval (ISMIR).
    [3] CKIP, http://ckipsvr.iis.sinica.edu.tw/
    [4] Logan, B., Kositsky, A., & Moreno, P. (2004, June). Semantic analysis of song lyrics. In Multimedia and Expo, 2004. ICME'04. 2004 IEEE International Conference on (Vol. 2, pp. 827-830). IEEE.
    [5] Mahedero, J. P., MartÍnez, Á., Cano, P., Koppenberger, M., & Gouyon, F. (2005, November). Natural language processing of lyrics. In Proceedings of the 13th annual ACM international conference on Multimedia (pp. 475-478). ACM.
    [6] Fell, M., & Sporleder, C. (2014). Lyrics-based Analysis and Classification of Music. In COLING (Vol. 2014, pp. 620-631).
    [7] Cai, R., Zhang, C., Wang, C., Zhang, L., & Ma, W. Y. (2007, September). Musicsense: contextual music recommendation using emotional allocation modeling. In Proceedings of the 15th ACM international conference on Multimedia (pp. 553-556). ACM.
    [8] Extended-HowNet, http://ehownet.iis.sinica.edu.tw/index.php
    [9] Chen, C. M., Tsai, M. F., Liu, J. Y., & Yang, Y. H. (2013, October). Using emotional context from article for contextual music recommendation. In Proceedings of the 21st ACM international conference on Multimedia (pp. 649-652). ACM.
    [10] Chen, K. J., Huang, S. L., Shih, Y. Y., & Chen, Y. J. (2005a). Extended-HowNet- A Representational Framework for Concepts. In Processing of OntoLex 2005 - Ontologies and Lexical Resources IJCNLP-05 Workshop, Jeju Island, South Korea.
    [11] Huang, S. L., Chung, Y. S., & Chen, K. J. (2008). E-HowNet- an Expansion of HowNet. In Proceedings of 1st National HowNet Workshop, Beijing, China.
    [12] CKIP Group (2013 reversion), Lexical Semantic Representation and Semantic Composition- An Introduction to E- How Net, CKIP Technical Report.
    [13] Wen, P. C., Tsai, Y. L., & Tsai, R.T. (2015). 基於 Word2Vec 詞向量的網路情緒文和流行音樂媒合方法之研究(Matching Internet Mood Essays with Pop-Music Based on Word2Vec)[In Chinese]. on Computational Linguistics and Speech Processing ROCLING XXVII (2015), 167.
    [14]陳學志, 詹雨臻, & 馮彥茹. (2013). 台灣地區華人情緒與相關心理生理資料庫─ 中文情緒隱喻的刺激常模. 中華心理學刊, 55(4), 525-553.
    [15] Blei, D. M., Ng, A. Y., & Jordan, M. I. (2003). Latent dirichlet allocation. Journal of machine Learning research, 3(Jan), 993-1022.
    [16] Ramage, D., Hall, D., Nallapati, R., & Manning, C. D. (2009, August). Labeled LDA: A supervised topic model for credit attribution in multi-labeled corpora. In Proceedings of the 2009 Conference on Empirical Methods in Natural Language Processing: Volume 1-Volume 1 (pp. 248-256). Association for Computational Linguistics.
    [17] Sahlgren, M., & Cöster, R. (2004, August). Using bag-of-concepts to improve the performance of support vector machines in text categorization. In Proceedings of the 20th international conference on Computational Linguistics (p. 487). Association for Computational Linguistics.
    [18] Wang, F., Wang, Z., Li, Z., & Wen, J. R. (2014, November). Concept-based short text classification and ranking. In Proceedings of the 23rd ACM International Conference on Conference on Information and Knowledge Management (pp. 1069-1078). ACM.
    [19] Hu, X., & Downie, J. S. (2010, August). When Lyrics Outperform Audio for Music Mood Classification: A Feature Analysis. In ISMIR (pp. 619-624).
    [20] Kleedorfer, F., Knees, P., & Pohle, T. (2008, September). Oh Oh Oh Whoah! Towards Automatic Topic Detection In Song Lyrics. In Ismir (pp. 287-292).
    [21] Sasaki, S., Yoshii, K., Nakano, T., Goto, M., & Morishima, S. (2014, October). LyricsRadar: A Lyrics Retrieval System Based on Latent Topics of Lyrics. In Ismir (pp. 585-590).
    [22] 楊帆. 淺析中國流行歌曲歌詞美學特性中的文學性(On the characteristics of Chinese pop music lyrics in literary aesthetics).[In Chinese] 音樂大觀, 2014, 9: 219-219.
    [23] Fell, M., & Sporleder, C. (2014). Lyrics-based Analysis and Classification of Music. In COLING (Vol. 2014, pp. 620-631).
    [24] Choi, K., Lee, J. H., & Downie, J. S. (2014, September). What is this song about anyway?: Automatic classification of subject using user interpretations and lyrics. In Proceedings of the 14th ACM/IEEE-CS Joint Conference on Digital Libraries (pp. 453-454). IEEE Press.
    [25] Choi, K., Lee, J. H., Hu, X., & Downie, J. S. (2016). Music subject classification based on lyrics and user interpretations. Proceedings of the Association for Information Science and Technology, 53(1), 1-10.
    [26] Song, Y., Dixon, S., & Pearce, M. (2012, June). A survey of music recommendation systems and future perspectives. In 9th International Symposium on Computer Music Modeling and Retrieval (Vol. 4).
    [27] Oudenne, A. M., Swarthmore, P. A., & Chasins, S. E. Identifying the Emotional Polarity of Song Lyrics through Natural Language Processing.
    [28] Hyung, Z., Lee, M., & Lee, K. (2012, October). Music recommendation based on text mining. In Proceedings of the 2nd international conference on advances in information mining and management (pp. 129-134).
    [29] Kaminskas, M., & Ricci, F. (2012). Contextual music information retrieval and recommendation: State of the art and challenges. Computer Science Review, 6(2), 89-119.
    [30] de Carvalho, A., & Freitas, A. (2009). A tutorial on multi-label classification techniques. Foundations of Computational Intelligence Volume 5, 177-195.
    [31] Boutell, M. R., Luo, J., Shen, X., & Brown, C. M. (2004). Learning multi-label scene classification. Pattern recognition, 37(9), 1757-1771.

    無法下載圖示 校內：2022-07-25公開
    校外：不公開
    電子論文尚未授權公開，紙本請查館藏目錄
    QR CODE