CITIES: Contextual Inference of Tail-Item Embeddings for Sequential Recommendation

Abstract

Sequential recommendation techniques provide users with product recommendations fitting their current preferences by handling dynamic user preferences over time. Previous studies have focused on modeling sequential dynamics without much regard to which of the best-selling products (i.e., head items) or niche products (i.e., tail items) should be recommended. We scrutinize the structural reason for why tail items are barely served in the current sequential recommendation model, which consists of an item-embedding layer, a sequence-modeling layer, and a recommendation layer. Well-designed sequence-modeling and recommendation layers are expected to naturally learn suitable item embeddings. However, tail items are likely to fall short of this expectation because the current model structure is not suitable for learning high-quality embeddings with insufficient data. Thus, tail items are rarely recommended. To eliminate this issue, we propose a framework called CITIES, which aims to enhance the quality of the tail-item embeddings by training an embedding-inference function using multiple contextual head items so that the recommendation performance improves for not only the tail items but also for the head items. Moreover, our framework can infer new-item embeddings without an additional learning process. Extensive experiments on two realworld datasets show that applying CITIES to the state-of-the-art methods improves recommendation performance for both tail and head items. We conduct an additional experiment to verify that CITIES can infer suitable new-item embeddings as well.

FPAdaMetric: False-Positive-Aware Adaptive Metric Learning for Session-Based Recommendation

“…Evaluation Metric We choose HR@K (Hit Ratio), MRR@K (Mean Reciprocal Rank) and NDCG@K (Normalized Discounted Cumulative Gain) for our evaluation metrics. These metrics are widely used in many of previous works (Jang et al 2020;Lv et al 2020). We also compare the results using different K ∈ {10, 50, 100} to verify robustness in various scenarios.…”

“…In this section, we formulate a general SBRs' model into an equation with trainable parameters of θ. A SBRs' model can be divided into 1) an embedding layer with θ e , 2) a sequential layer with θ seq , and 3) a recommendation layer (Jang et al 2020;Lv et al 2020). The trainable parameters are consist of the followings: θ = {θ e , θ seq , θ concat (optional)}, where θ e = θ i e , θ u e .…”

“…STAMP, GRU4REC) are reported in the appendix. Due to the low performance of the traditional non-deep learning methods(Jang et al 2020;Qiu et al 2020) -Sequence Popularity (S-POP) or First-order Markov Chain (FOMC)-are excluded in the baseline.…”

Freudian and Newtonian Recurrent Cell for Sequential Recommendation

“…2. Generally, a sequential recommendation model F is composed of three layers: item embedding, sequence modelling, and recommendation layers [28]. The item embedding layer encodes items into a low-dimensional real-valued latent space to measure the similarity between items.…”

“…These vectors are very sparse, and the similarity between items is difficult to measure. Recent works [8,11,28,32] employ an embedding layer to resolve these limitations, and this study also uses an item embedding layer. The item embedding layer f : N → R du projects the item index onto a d u -dimensional real-valued dense vector as…”