Start Date
12-18-2013
Description
The rapid development of technology promotes the vast expansion of new items in many domains of consumer products. Problem occurs when the new items are continuously added but cannot get reached by the consumers. Many existing recommender systems work well only for well-known items with sufficient ratings but fail to discover new items, and content-based approaches suffer from insufficient item features. In this paper, we show that critic reviews of the items can be used to boost new item recommendation. We propose a scalable framework that incorporates the topics inferred from the critic reviews into the recommendation process by employing topic modeling and non-negative matrix factorization. The results of our experiment show that our proposed method is able to generate high quality new item recommendations which are not supported by many state-of-the-art methods, and also outperforms the state-of-the-art methods in recommending existing items.
Recommended Citation
Xu, Xiaoying; Dutta, Kaushik; and Datta, Anindya, "Using Critic Reviews to Boost New Item Recommendation" (2013). ICIS 2013 Proceedings. 12.
https://aisel.aisnet.org/icis2013/proceedings/KnowledgeManagement/12
Using Critic Reviews to Boost New Item Recommendation
The rapid development of technology promotes the vast expansion of new items in many domains of consumer products. Problem occurs when the new items are continuously added but cannot get reached by the consumers. Many existing recommender systems work well only for well-known items with sufficient ratings but fail to discover new items, and content-based approaches suffer from insufficient item features. In this paper, we show that critic reviews of the items can be used to boost new item recommendation. We propose a scalable framework that incorporates the topics inferred from the critic reviews into the recommendation process by employing topic modeling and non-negative matrix factorization. The results of our experiment show that our proposed method is able to generate high quality new item recommendations which are not supported by many state-of-the-art methods, and also outperforms the state-of-the-art methods in recommending existing items.