Research On Book Retrieval And Recommendation Techniques

The rapid development of modern society has significantly boosted the book publishingindustry and made the collection of books in most libraries. However, the book retrieval andrecommendation techniques based on strict text matching fails to match this flourish and providegood-quality services to readers. Tu proposed in his thesis the BookRank algorithm, an effectivesolution which takes advantage of the book recommendation information provided by onlinebook shops. However, it cannot deal with those books that are not on the shops’ list, i.e. thosethat are not covered by BookRank.To solve the recommendation problem of these legacy books, this paper presents aBookValue algorithm which aims to unveil the relevance between books by analyzing theirborrow records. Recommendations can therefore be made for these books. The underlyingphilosophy of BookValue is very similar to that of BookRank except that it takes into concernthe reader’s expertise in related domain and their values in book recommendation. A prototypesystem has been developed based on J2EE enterprise architecture and SQL Server2005RDBMS.Preliminary experimental results show that the above system can recommend those legacybooks of the library to users who may find them useful.