Embedding unstructured side information in product recommendation

作者：

Highlights：

• We propose employing a variety of unstructured content to improve recommendation.

• Unstructured product and user content are modeled through word embedding approaches.

• Both graph based link prediction and feature based matrix factorization are extended.

• Our feature based matrix factorization model improves ranking and rating prediction.

摘要

•We propose employing a variety of unstructured content to improve recommendation.•Unstructured product and user content are modeled through word embedding approaches.•Both graph based link prediction and feature based matrix factorization are extended.•Our feature based matrix factorization model improves ranking and rating prediction.

论文关键词：Matrix factorization,User and product embeddings,Recurrent neural networks,Recommender systems

论文评审过程：Received 1 February 2017, Revised 3 August 2017, Accepted 3 August 2017, Available online 10 August 2017, Version of Record 5 September 2017.

论文官网地址：https://doi.org/10.1016/j.elerap.2017.08.001