DEIM Forum 2012 A1-3 525 8577 1 1 1 E-mail: oku@fc.ritsumei.ac.jp, fhattori@is.ritsumei.ac.jp Kenta OKU and Fumio HATTORI College of Information Science and Engineering, 1 1 1 Nojihigashi, Kusatsu-city, Shiga, 525 8577, Japan E-mail: oku@fc.ritsumei.ac.jp, fhattori@is.ritsumei.ac.jp 1. [1] [2] 3 [3] 3 Horace Walpole [4] Walpole [4] 2 [5] 2002
(a) (b) (c) a) b) a) b) 1 1 Fig. 1 Interface of fusion-based recommender system. & 2. 3. 4. 5. 2. Herlocker [1] Ziegler [6] [7] Sarwar [8]
Hijikata [9] Murakami [2] Hijikata [9] Murakami [2] 3. 1 (a) (b) (c) 1 2 & 3. 1 3. 2 3. 3 3. 1 2 Fig. 2 Concept of fusion based on base-item and material-item. Table 1 1 Principal attributes of book data. 1 API 2 2011 12 27 2012 2 10 667, 218 1 book book-phrase(isbn, phrase, idf) book-author(isbn, author) book-genre(isbn, genre id) book-phrase book.title book.sub title 3. 1. 1 book-author book-genre ID 4 800 ID 3. 1. 1 book.title book.sub title 3 1 http://books.rakuten.co.jp/book/ 2 API http://webservice.rakuten.co.jp/api/booksbooksearch/ 3 http://chasen.naist.jp/hiki/chasen/
2 Table 2 Search processing by each button. IDF IDF w i IDF i IDF i = log N (1) n i N IDF N = 667, 218 n i book.title book.sub title w i 3. 2 1 (a), (b), (c) (a) N 1 I, II, III 2 θ 4. θ = 1000 (b) 3 Fig. 3 Example of each fusion. & 3. 3 3 I, II, III (c) (b) 3. 3 3. 2(b) & 3 booka bookb book a ) phrase phrase booka booka.phraselist bookb bookb.phraselist
book.title book.sub title M I 3(a) booka bookb booka bookb book.title book.sub title phrase phrase Algorithm1 searchbyphrasephrase(phrasea, phraseb, count) book book.title book.sub title phrasea phraseb count booka.phraselist bookb.phraselist IDF IDF b ) phrase genre booka booka.phraselist book.title book.sub title bookb bookb.genre idlist book.genre id M II 3(b) booka bookb booka book.title book.sub title book.genre id bookb.genre id - book book.title book.sub title phrasea book.genre id genre idb M M genre idb - 2 c ) phrase author booka booka.phraselist book.title book.sub title bookb bookb.authorlist book.author M III 3(c) booka bookb booka book.title book.sub title book.author bookb.author Algorithm 1 phrase phrase 1: booklist = [] 2: for phrasea in booka.phraselist do 3: for phraseb in bookb.phraselist do 4: list = searchbyphrasephrase(phrasea, phraseb, M + 2) 5: booklist.add(list) 6: if booklist.size > M + 2 then 7: break 8: end if 9: end for 10: if booklist.size > M + 2 then 11: break 12: end if 13: end for 14: booklist.remove(booka) 15: booklist.remove(bookb) 16: return booklist book book.title book.sub title phrasea book.author authorb M 4. 3. Java Processing 3. 1 MySQL 4. 1 9 8 1 20 23 Amazon 1-3 1 3 3 (1) 3 (2) (3) 3 (4) 0 (5) 3 3
3 Table 3 Questions for recommended books. 4 Table 4 Questions for overall recommender systems. (6) 3 (7) 4. 2 3 Q1 3 {3: 2:1: } Q2 Q5 5 {5: 4: 3:2:1: } Q5 Google Yahoo! 4 5 Pu [10] 4 4. 2 Amazon 4 (A- RS)Amazon (A-Rank)Amazon 2 (F-P) (F-G) 2 (A-RS) Amazon Amazon Amazon 4 amazon.co.jp http://www.amazon.co.jp/ Amazon [11] (A-Rank) Amazon Amazon (F-P) 3. 2 3. 3 (F-G) 1 3 3 3 4. 3 4. 3. 1 4 Q1 Q5 3 4 Q5 F-P F-G A-RS A-Rank F-P A-RS A-Rank F-G A-RS A-Rank t 5% F-P F-G 1.
Q1 Q5 Q3 4 Q1 3 Q5 4 Q3 4 1 0 t F-G A-RS A-Rank 5% 1. Q2 Q3 4 Q2 2 Q3 4 1 0 t F-P A-RS F-G A-RS A-Rank 1% F-P A-Rank 5% A-RS A-Rank 4 Fig. 4 Separate evaluation of sub recommended book. 5 Fig. 5 Evaluation of overall system. 4. 3. 2 5 A-RS A-Rank Q1 Q11 4 Q11 A-RS A-Rank 5% 1. Q6: A-RS A-Rank A-RS A-Rank
Pu [10] Q2. Q3. Q8 2 2 4 2 Q9 2 2 Q4: 2 Q4 Q8 Q4 Q9 0.781, 0.877 Q4 4 Q8 Q9 5. Amazon B 23700132 [1] J.L. Herlocker, J.A. Konstan, L.G. Terveen, and J.T. Riedl. Evaluating collaborative filtering recommender systems. ACM Transactions on Information Systems (TOIS), Vol. 22, No. 1, pp. 5 53, 2004. [2] Tomoko Murakami, Koichiro Mori, and Ryohei Orihara. Metrics for evaluating the serendipity of recommendation lists. New Frontiers in Artificial Intelligence, pp. 40 46, 2008. [3] Elizabeth Jamison Hodges ( ), ( ) et al.., 2004. [4] ( ) and ( ).., 2007. [5] Royston M. Roberts ( ) and ( ).., 1993. [6] Cai-Nicolas Ziegler, Georg Lausen, and Lars Schmidt- Thieme. Taxonomy-driven computation of product recommendations. In Proceedings of the Thirteenth ACM conference on Information and knowledge management - CIKM 04, p. 406, New York, New York, USA, 2004. ACM Press. [7] C.N. Ziegler, S.M. McNee, J.A. Konstan, and Georg Lausen. Improving recommendation lists through topic diversification. In Proceedings of the 14th international conference on World Wide Web, pp. 22 32, New York, New York, USA, 2005. ACM. [8] Badrul Sarwar, George Karypis, Joseph Konstan, and John Reidl. Item-based collaborative filtering recommendation algorithms. In Proceedings of the 10th international conference on World Wide Web, pp. 285 295, New York, New York, USA, 2001. ACM. [9] Y. Hijikata, T. Shimizu, and S. Nishida. Discovery-oriented collaborative filtering for improving user satisfaction. In Proceedings of the 13th international conference on Intelligent user interfaces, pp. 67 76. ACM, 2009. [10] P. Pu and L. Chen. A User-Centric Evaluation Framework for Recommender Systems. In RecSys 11:Proceedings of the fifth ACM conference on Recommender systems, pp. 157 164. ACM, 2011. [11] G. Linden, B. Smith, and J. York. Amazon.com recommendations: Item-to-item collaborative filtering. In IEEE Internet Computing, vol. 7, no. 1, pp. 76 80, 2003.