《利用社交网络的旅游推荐系统》由会员分享，可在线阅读，更多相关《利用社交网络的旅游推荐系统（8页珍藏版）》请在金锄头文库上搜索。
1、With a Little Help From My Friends: Generating Personalized Book Recommendations Using Data Extracted from a Social WebsiteMaria Soledad PeraYiu-Kai Ng Computer Science Department Brigham Young University Provo, Utah, U.S.A.AbstractWith the large amount of books available nowadays, users are overwhe
2、lmed with choices when they attempt tofind books of interest. While existing book recommendationsystems, which are based on either collaborative filtering, content-based, or hybrid methods, suggest books (among the millions available) that might be appealing to the users, their recommendationsare no
3、t personalizedenoughto meet users expectations due to their collective assumption on group preference and/or exact content matching, which is a failure.To address this problem, we have developed PReF, a Personalized Recommender that relies on Friend- ships established by users on a social website, s
4、uch as Li- braryThing, to make book recommendations tailored to in- dividual users. In selecting books to be recommended to a user U, who is interested in a book B, PReF (i) considers books belongedto Us friends, (ii) applies word-correlation factors to disclose books similar in contents to B, (iii)
5、 de- pends on the ratings given to books by Us friends to iden- tify highly-regardedbooks,and(iv)determines howreliable individual friends of U are in providing books from their own catalogs (that are similar in content to B) to be recom- mended. We have conducted an empirical study and veri-fied th
6、at (i) relying on data extracted from social websites improves the effectiveness of book recommenders and (ii) PReF outperforms the recommenders employed by Ama- zon and LibraryThing.1IntroductionIn recent years social websites, such as Facebook(.com), Twitter(.com), YouTube(.com), and Delicious(.co
7、m), have become increasinglypopular 9. These sites introducenew user-generated data and metadata, such as ratings, social connections,andtags1, whichprovidearichsourceofinfor-1Tags are user-defined keywords that describe the content of an item.mationtoinferusersinterests andpreferences. Thesekinds o
8、f information are unique and valuable for making recom- mendations on books, movies, news articles, etc., which have been examined in 2, 3, 9.Newly-developed rec- ommenders, such as 9, 16, 22, incorporate data extracted from social websites to increase the quality of tag, news articles, and book rec
9、ommendations. Book recommenders have been adopted by online shopping companies, social websites, and digital libraries, to name a few, to further facilitate their users knowledge acquisition process by of- fering alternative choices (among the millions available) of books they are likely interested
10、in. While suggestions pro- vided by existing book recommenders can introduce users to books that they are not aware of, these recommendersare not personalized enough to achieve their design goals 10. It is imperative to develop personalized recommenders thatprovide finer suggestions pertinent to ind
11、ividual users in- terests or preferences. To the best of our knowledge, there are no recommendation systems that simultaneously con- sider users relationships, along with user-generated data extracted from a social website, to recommend books. In this paper, we introduce PReF, a personalized book re
12、commendation system that depends on friendships estab- lished among users in a social website, which is Library- Thing2in our case, to generate valuable book recommenda- tions tailored to individual users interests. PReF locates, among the books bookmarked by Us friends on a social website, the ones
13、 that are similar in content to a given book B that U is interested in. Hereafter, PReF ranks the candi- date books to be recommended by considering not only the content similarity between each candidate book CB and B, but also the ratings assigned to CB by Us friends, and the reliability of each of
14、 Us friends. PReF is an elegant and unique system that relies on (i) relationships established between a user and other mem-2LibraryThing(.com) was founded in 2006 for aiding users in cata-loging and referencing books. LibraryThing users can rate and review books, add tags to books to describe their
15、 contents, and establish friend- ships, i.e., bi-directional relationships, with other LibraryThing users.1bers of a social website, since as stated in 2, the quality of recommendations given to a user U is improved by con- sidering opinions of other users whom U trusts, (ii) ratings provided by use
16、rs of a social site, which aid in identifying highly-regarded books a user might be interested in, and (iii) word-correlation factors 12, which detect books sim- ilar in content, even if they are described using analogous, but not the same, tags, to generate personalized book rec- ommendations. In addition, PReF can perform the recom- mendationtask with data extractedfrom any social website, provided that users relationships, book tags, and book rat- ings can be obtained
