結合社會性標籤與文獻內容於個人化學術文章推薦

網路普及帶動資訊快速流通，查詢資料變得簡單且方便，但如何在這些資料中獲得有用且符合需求的資訊卻成為最大的困難點。在學術領域中，使用期刊文獻資料庫進行學術文章搜索一向是學者專家所仰賴的方法，但學者往往要針對搜尋出來的結果進行資料過濾，面對日積月累所產生出來的學術資料，必須要制定一個方法來改善這樣的問題。學術文章推薦是近年來學術研究的熱門議題，過往針對學術文章推薦研究上，學者普遍只使用學術文章內所擁有的屬性，如：標題、摘要、關鍵字、作者名稱以及參考文獻標題等進行推薦。本研究提出了一種利用擴充文章字詞的推薦方法，運用傳統內容導向式推薦中未考慮的文章屬性：參考文獻內容以及相同標籤的文章內容等所產生的關鍵字集來補強原始文章關鍵字集的不足，降低因字詞缺乏而產生的資料稀疏性問題。且運用層級分析法制定出文章屬性間的權重值，將不同文章屬性間的相似度值進行線性合併，最後產生總相似度值進行推薦。運用本研究所提出之文章推薦方法，便能解決學者在搜尋學術文章資料上的時間成本，減少繁瑣的文章過濾程序。在實驗的部分，本研究針對五個知名期刊進行文章收集並區分為三個不同實驗資料，且自行建構實驗環境進行測驗，使用三個評量指標：成對比較法(Pair Match)、命中率(Hit Rate)以及評分法(Score Match)等方法驗證推薦系統效果。實驗結果顯示，本研究所提出之推薦方法，能夠精準推薦讀者所優先喜好的文章，優於傳統只考慮文章原有屬性的推薦方法，以證明本研究之推薦方法是具有較佳的文章推薦效果。

關鍵字

社會性標籤；學術文章推薦；推薦系統；文字探勘

並列摘要

The rise of digital libraries makes the acquisition of academic resources become simpler, but a large number of academic articles lead to the difficulties of searching for researchers. The existing retrieval tools of academic articles mainly rely on the search of keywords, and users still have to filter the search results. In recent years, the techniques of document recommendation (DR) are applied in the recommendation of academic articles, which gives the automated and personalized recommendation services according to user's preferences. However, there are still two potential problems remained. The first is that the recommendation of academic documents has the data sparsity problem. The second is that the contributions of an academic paper are often constructed on the bases of previous studies and therefore the academic articles may not include complete research background and relevant knowledge for their research problems. For the abovementioned two problems, this study attempts to improve the recommendation performance of academic articles through the consideration of external documental knowledge. Specifically, we use the social tagging and the content of reference articles to extend document vectors. In addition, the degrees of importance of the seven attributes are not same at all. This study adopts the analytic hierarchy process (AHP) method to determine the weights among the seven article attributes. In experimental evaluation, we develop a recommendation system, which make recommendations for users by several research document techniques. We uses three evaluation indicators : Pair Match, Hit Rate and Score Match Method. Experimental results show that our recommendation method can recommend users priority preferences document. In the past, researcher only consider the original document attributes. In this study, our recommended method has better performance than traditional recommendation method.

並列關鍵字

Social Tagging ； Research document Recommendation ； Recommendation Systems ； Text Mining

參考文獻

Ahlgren, P., & Colliander, C. (2009). Document document similarity approaches and science mapping: Experimental comparison of five approaches. Journal of Informetrics, 3(1), 49-63. doi: 10.1016/j.joi.2008.11.003

Cao, Y., & Li, Y. (2007). An intelligent fuzzy-based recommendation system for consumer electronic products. Expert Systems with Applications, 33(1), 230-240. doi: 10.1016/j.eswa.2006.04.012

Chen, Y., Tsai, F. S., & Chan, K. L. (2008). Machine learning techniques for business blog search and mining. Expert Systems with Applications, 35(3), 581-590. doi: 10.1016/j.eswa.2007.07.015

Chen, Y. L., Wei, J. J., Wu, S. Y., & Hu, Y. H. (2006). A similarity-based method for retrieving documents from the SCI/SSCI database. Journal of information science, 32(5), 449-464. doi: 10.1177/0165551506065814

Cheng, T. H., & Wei, C. P. (2008). A clustering-based approach-for integrating document-category hierarchies. IEEE Transactions on Systems Man and Cybernetics Part a-Systems and Humans, 38(2), 410-424. doi: 10.1109/TSMCA.2007.914758

國際替代計量

結合社會性標籤與文獻內容於個人化學術文章推薦

未授權

主題瀏覽