User Profiling from Reviews for Accurate Time-Based Recommendations.

摘要: Recommender systems are a valuable way to engage users in system, increase participation and show them resources they may not have found otherwise. One significant challenge is that user interests change over time certain items an inherently temporal aspect. As result, recommender system should try take into account the time-dependant user-item relationships. However, aspects of profile always be explicitly available so we need infer this information from resources. Product reviews on sites, such as Amazon, represent data source understand why someone bought item potentially who for. This can then used construct dynamic profile. In paper, demonstrate utilising extract \textit{age category preference} users, leverage feature generate time-dependent recommendations. Given predictable yet shifting nature age time, that, recommendations generated using aspect lead higher accuracy compared with techniques state art. Mining temporally related content enable go beyond finding similar or predict future user.