Modeling User Fatigue for Sequential Recommendation

TL;DR

This work addresses user fatigue in sequential recommendation by introducing FRec, a model that integrates fine grained interest aware similarity features, fatigue aware fusion for long term interests, a fatigue gated recurrent unit for short term dynamics, and fatigue supervised contrastive learning via sequence augmentation. The approach jointly optimizes a standard recommendation loss with a fatigue aligned contrastive loss, leading to consistent improvements on public datasets and large scale online deployments. Key contributions include the interest aware similarity matrix, fatigued cross interaction fusion, fatigue driven short term evolution, and explicit fatigue supervision for representation learning, all validated through extensive offline and online experiments. The results demonstrate not only higher accuracy metrics (e.g., AUC and GAUC gains) but also tangible fatigue reduction in real user interactions, underscoring the practical impact for improving user experience in recommender systems.

Abstract

Recommender systems filter out information that meets user interests. However, users may be tired of the recommendations that are too similar to the content they have been exposed to in a short historical period, which is the so-called user fatigue. Despite the significance for a better user experience, user fatigue is seldom explored by existing recommenders. In fact, there are three main challenges to be addressed for modeling user fatigue, including what features support it, how it influences user interests, and how its explicit signals are obtained. In this paper, we propose to model user Fatigue in interest learning for sequential Recommendations (FRec). To address the first challenge, based on a multi-interest framework, we connect the target item with historical items and construct an interest-aware similarity matrix as features to support fatigue modeling. Regarding the second challenge, built upon feature cross, we propose a fatigue-enhanced multi-interest fusion to capture long-term interest. In addition, we develop a fatigue-gated recurrent unit for short-term interest learning, with temporal fatigue representations as important inputs for constructing update and reset gates. For the last challenge, we propose a novel sequence augmentation to obtain explicit fatigue signals for contrastive learning. We conduct extensive experiments on real-world datasets, including two public datasets and one large-scale industrial dataset. Experimental results show that FRec can improve AUC and GAUC up to 0.026 and 0.019 compared with state-of-the-art models, respectively. Moreover, large-scale online experiments demonstrate the effectiveness of FRec for fatigue reduction. Our codes are released at https://github.com/tsinghua-fib-lab/SIGIR24-FRec.

Figures (9)

• Figure 1: (a) An illustration demo to show how we measure EVTR along with user consumption. Color represents the video category, and blue is the target category to count the number of effective views. (b) The trend of EVTR with increasing consumption of videos with the same category. We include the EVTR of the video with the target category (same) and videos with other categories close to it (other).
• Figure 2: The framework of FRec.
• Figure 3: Illustration of the calculation of interest-aware similarity, represented by the length of the red line.
• Figure 4: Illustration of the sequence augmentation to obtain fatigue signals. There is more fatigue if some items in the sub-sequence $\hat{S}_u$ are replaced by the target item.
• Figure 5: Ablation study of key modules. Error bar denotes 95% confidence intervals for five experiments.