Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/51578
標題: 提升圖書館個人本體論相似度計算速度之研究
An Efficient Collaborative Filtering Technique for Personal Ontology Model in Library Recommender System
作者: 廖宜恩
關鍵字: 合作式過濾
資訊科學--軟體
Personal Ontology Model
結構相似度
個人本體論
推薦系統
應用研究
Recommender Systems
Collaborative Filtering
Content-Based Filtering
Edit Distance of Trees
摘要: The research on recommender systems has drawn a lot of public attentions since the announcement of Netflix competition with the prize of 1 million US dollars in October 2006. The essential problem underlying the Netflix competition is the efficiency and accuracy problem of a recommender system with the magnitude of Netflix, which is a movie rental company having 480,189 customers and 17,770 movies.In 2006, we proposed a novel model called Personal Ontology Model for library recommender system. This model was implemented in the Library of National Chung Hsing University for recommending Chinese collections in 2007. We then upgraded the system, called PORE (Personal Ontology Recommender), by adding collaborative filtering techniques and including English collections recommendation.Although the PORE system is in operation, it still suffers from efficiency problem. The motivation of this research proposal is not only to design a more accurate and effective recommendation system for libraries, but also to eliminate many of the weaknesses found in the existing recommender systems. In the proposed approach, we will define a similarity function for personal ontology model based on the edit distance of trees and propose a new scheme for efficiently computing the similarities among library patrons.
推薦系統的研究,近年來受到高度的重視。一個準確的推薦系統,能夠提高購物網站的營業額或者提升圖書館資訊服務的品質。推薦系統的作法,可以分為以內容為基礎(content-based),及以合作式過濾為基礎(collaborative filtering) 的推薦。以合作式過濾為基礎的系統,必須能夠快速的尋找到擁有類似喜好的使用者。當系統找出相似的使用者,便可以依照使用者的行為紀錄(購買紀錄或是借閱紀錄),找出這個族群喜好的物品,進而推薦該族群使用者尚未購買或借閱的物品。而將使用者的喜好特徵,依照領域知識,組織成一個有系統的結構,我們稱之為個人本體論。使用個人本體論來表達個人喜好,可以同時表達喜好的領域分類及喜好間的語意意涵,因此較諸傳統使用關鍵字的方式要來得有效。但是,在比較個人本體論的相似性時,必須做結構相似度的比對,其速度卻遠較傳統的關鍵字比對的來得慢。因此如何改善個人本體論相似度計算的速度,是一個非常重要的議題。本研究計劃的目標,首先將設計一個適用於圖書館本體論的相似度演算法。此外,我們將著重在如何提升個人本體論相似度計算的速度。我們將分成兩個情境來探討提升計算速度的方法。第一:當系統可以接受概略解(Approximate Solution)時。第二:當系統要求精確解(Exact Solution)時。對這兩個情況可用的技巧,本研究團隊目前已討論出初步的構想。在概略解方面,我們計劃先過濾喜好度過低的特徵分類;在精確解方面,我們將設計一個符合三角不等式概念的衡量基準,最佳的情況將能大量降低系統的運算時間。本計劃將更深入研究有關計算樹狀結構edit distance的有效率演算法,同時將會把本研究所提出的技巧,實做在中興大學的圖書館個人化推薦系統。本研究的執行,預計將在學術研究、實務應用及學生能力的增進上,都帶來極大的助益。
URI: http://hdl.handle.net/11455/51578
其他識別: NSC98-2221-E005-083
文章連結: http://grbsearch.stpi.narl.org.tw/GRB/result.jsp?id=1899080&plan_no=NSC98-2221-E005-083&plan_year=98&projkey=PB9808-0478&target=plan&highStr=*&check=0&pnchDesc=%E6%8F%90%E5%8D%87%E5%9C%96%E6%9B%B8%E9%A4%A8%E5%80%8B%E4%BA%BA%E6%9C%AC%E9%AB%94%E8%AB%96%E7%9B%B8%E4%BC%BC%E5%BA%A6%E8%A8%88%E7%AE%97%E9%80%9F%E5%BA%A6%E4%B9%8B%E7%A0%94%E7%A9%B6
Appears in Collections:資訊科學與工程學系所

文件中的檔案:

取得全文請前往華藝線上圖書館

Show full item record
 
TAIR Related Article
 
Citations:


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.