Sequential Recommender Systems: Challenges, Progress and Prospects
Shoujin Wang, Liang Hu, Yan Wang, Longbing Cao, Michael Sheng and Mehmet Orgun, IJCAI2019 survey paper

The emerging topic of sequential recommender systems (SRSs) has attracted increasing attention in
recent years. Different from the conventional recommender systems (RSs) including collaborative filtering and content-based filtering, SRSs try to understand and model the sequential user behaviors, the interactions between users and items, and the evolution of users’ preferences and item popularity over time. SRSs involve the above aspects for more precise characterization of user contexts, intent and goals, and item consumption trend, leading to more customized and dynamic recommendations. In this paper, we provide a systematic review on SRSs. We first present the characteristics of SRSs, and then summarize and categorize the key challenges in this research area, followed by the corresponding research progress consisting of the most recent and representative developments on this topic. Finally, we discuss the important research directions in this vibrant and quickly expanding area.

Modeling Multi-Purpose Sessions for Next-Item Recommendations via Mixture-Channel Purpose Routing Networks
Shoujin Wang, Yan Wang, Liang Hu, Michael Sheng and Mehmet Orgun, Longbing Cao, IJCAI2019 paper

A session-based recommender system (SBRS) recommends the next item according to the dependencies between items in a session context. Most existing SBRSs assume an implicit contextual goal to correlated items inside a session. However, this may not always be true in reality. In fact, a session is frequently composed of multiple subsets of items w.r.t. different purposes (e.g., breakfast and decoration). More specifically, items (e.g., bread and milk) in the same subset have strong purpose-specific dependencies whereas items (e.g., bread and vase) from different subsets have much weaker or even no dependency due to the difference of purposes. Therefore, we propose a mixture-channel structure to accommodate the multi-purpose item subsets for more precisely describing a session context. Compared with traditional SBRSs, mixture channels naturally enable SBRSs to recommend more diverse items to satisfy different purposes. To this end, we design an effective mixture channel network (MCN) with a gating network to detect the purposes of each item and assign them into different channels. Moreover, a purpose-specific dependency gated recurrent neural network (PSD-GRU) is employed to model dependencies between items within each channel for a specific purpose. The experimental results show the superiority of MCN over the state-of-the-art methods in terms of both recommendation accuracy and diversity.