面向OPAC的非个性化图书推荐算法

Non-personalized Book Recommendation Algorithm oriented to OPAC

[目的/意义] OPAC书目检索系统用户处于非登录状态,系统无法获取用户个人信息,个性化推荐算法难以发挥作用,有必要探索非个性化推荐算法。[方法/过程]首先提出基于图书语义相似度的图书推荐算法,通过构建向量空间模型计算图书语义相似度,向读者推荐与当前浏览图书相似的图书;然后提出基于共借关系的图书推荐算法,向读者推荐借阅了当前浏览图书的读者还借阅过的其他图书;最后讨论了两种算法的融合策略。[结果/结论]选取10本图书作为推荐窗口,在复旦大学图书馆真实借阅数据集上进行实验,推荐成功率为20%。每5名读者中有1名读者能在推荐列表中发现自己后续会借阅的图书。

[Purpose/Significance] Non-personalized Book Recommendation Algorithm is necessary for OPAC book retrieval system,because users are always in the state of logging out. It is impossible to access user information,without which personalized recommendation algorithm cannot work efficiently. [Method/Process] Firstly,book recommendation algorithm based on semantic similarity was designed with vector space model recommending books similar to book browsing.Secondly,algorithm based on relation of same readers was introduced,which recommended books borrowed by readers who also borrowed the book that the user was browsing. Lastly,methods to merge two algorithm were discussed. [Results/Conclusion] With ten recommending books,the result of experiment on Fudan University library’s book borrowed datasets showed that success rate of algorithm was 20%,such that one in five readers could find at least one books that he would borrow in recommending book list.