學術產出-國科會研究計畫

文章檢視/開啟

書目匯出

Google ScholarTM

政大圖書館

引文資訊

TAIR相關學術產出

題名 流行音樂的詞曲搭配推薦系統
作者 沈錳坤
貢獻者 資訊科學系
關鍵詞 詞式結構分析;曲式結構分析;詞曲搭配;推薦系統
lyrics structure analysis; music structure analysis; lyrics music alignment; recommendation
日期 2013
上傳時間 8-十二月-2017 15:08:22 (UTC+8)
摘要 本計劃透過分析詞曲搭配的規則、研究流行音樂的詞曲搭配推薦技術及詞曲對齊演算法。此外,本研究研發並實作了以曲找詞的舊曲新詞推薦系統:透過使用者輸入歌曲,由系統自動查詢適合搭配的歌詞,並以歌聲模擬軟體試唱。本研究也可以延伸為以詞找曲的舊詞新曲推薦系統:透過使用者輸入歌詞,由系統自動查詢適合搭配的歌曲。本計劃的研究成果,透過雲端服務,提供舊曲新詞、舊詞新曲的音樂加值應用。
This project investigates the relationships between lyrics and music for Chinese popular song. We investigate the methods for music structure and lyrics structure analysis respectively and propose the alignment algorithm between a lyrics and a music object. For music structure analysis, we detect the music sentence based on Support Vector Machine and discover the music form by Non-negative Matrix Factorization of constructed Self Similarity Matrix of music sentences. For lyrics structure analysis, we considered four features, word count, Pinyin, Part-Of-Speech, and Word tone, to measure the similarity between lyrics sentences and construct the Self Similarity Matrix of lyrics sentences. The lyrics structure is analyzed by finding the cohesive blocks of the lyrics Self Similarity Matrix. Moreover, we proposed the dynamic programming algorithm to align a lyrics and a music object. We developed the lyrics recommendation system that recommends lyrics accompany well with the query music. Experimental results showed that the proposed approach performs well for Chinese popular song.
關聯 執行起迄:2013/08/01~2015/01/31
102-2221-E-004-005
資料類型 report
dc.contributor 資訊科學系zh_Tw
dc.creator (作者) 沈錳坤zh_TW
dc.date (日期) 2013en_US
dc.date.accessioned 8-十二月-2017 15:08:22 (UTC+8)-
dc.date.available 8-十二月-2017 15:08:22 (UTC+8)-
dc.date.issued (上傳時間) 8-十二月-2017 15:08:22 (UTC+8)-
dc.identifier.uri (URI) http://nccur.lib.nccu.edu.tw/handle/140.119/115091-
dc.description.abstract (摘要) 本計劃透過分析詞曲搭配的規則、研究流行音樂的詞曲搭配推薦技術及詞曲對齊演算法。此外,本研究研發並實作了以曲找詞的舊曲新詞推薦系統:透過使用者輸入歌曲,由系統自動查詢適合搭配的歌詞,並以歌聲模擬軟體試唱。本研究也可以延伸為以詞找曲的舊詞新曲推薦系統:透過使用者輸入歌詞,由系統自動查詢適合搭配的歌曲。本計劃的研究成果,透過雲端服務,提供舊曲新詞、舊詞新曲的音樂加值應用。zh_TW
dc.description.abstract (摘要) This project investigates the relationships between lyrics and music for Chinese popular song. We investigate the methods for music structure and lyrics structure analysis respectively and propose the alignment algorithm between a lyrics and a music object. For music structure analysis, we detect the music sentence based on Support Vector Machine and discover the music form by Non-negative Matrix Factorization of constructed Self Similarity Matrix of music sentences. For lyrics structure analysis, we considered four features, word count, Pinyin, Part-Of-Speech, and Word tone, to measure the similarity between lyrics sentences and construct the Self Similarity Matrix of lyrics sentences. The lyrics structure is analyzed by finding the cohesive blocks of the lyrics Self Similarity Matrix. Moreover, we proposed the dynamic programming algorithm to align a lyrics and a music object. We developed the lyrics recommendation system that recommends lyrics accompany well with the query music. Experimental results showed that the proposed approach performs well for Chinese popular song.en_US
dc.format.extent 3564002 bytes-
dc.format.mimetype application/pdf-
dc.relation (關聯) 執行起迄:2013/08/01~2015/01/31zh_TW
dc.relation (關聯) 102-2221-E-004-005zh_TW
dc.subject (關鍵詞) 詞式結構分析;曲式結構分析;詞曲搭配;推薦系統zh_TW
dc.subject (關鍵詞) lyrics structure analysis; music structure analysis; lyrics music alignment; recommendationen_US
dc.title (題名) 流行音樂的詞曲搭配推薦系統_TW
dc.type (資料類型) report