Xuanji Xiao

Xuanji Xiao, Ziyu He

Rank models play a key role in industrial recommender systems, advertising, and search engines. Existing works utilize semantic tags and user-item interaction behaviors, e.g., clicks, views, etc., to predict the user interest and the item hidden representation for estimating the user-item preference score. However, these behavior-tag-based models encounter great challenges and reduced effectiveness when user-item interaction activities are insufficient, which we called "the long-tail ranking problem". Existing rank models ignore this problem, but its common and important because any user or item can be long-tailed once they are not consistently active for a short period. In this paper, we propose a novel neighbor enhancement structure to help train the representation of the target user or item. It takes advantage of similar neighbors (static or dynamic similarity) with multi-level attention operations balancing the weights of different neighbors. Experiments on the well-known public dataset MovieLens 1M demonstrate the efficiency of the method over the baseline behavior-tag-based model with an absolute CTR AUC gain of 0.0259 on the long-tail user dataset.

Lin Wang, Weisong Wang, Xuanji Xiao et al.

Recommender systems are essential for modern content platforms, yet traditional behavior-based models often struggle with cold users who have limited interaction data. Engaging these users is crucial for platform growth. To bridge this gap, we propose leveraging the social-relation graph to enrich interest representations from behavior-based models. However, extracting value from social graphs is challenging due to relation noise and cross-domain inconsistency. To address the noise propagation and obtain accurate social interest, we employ a dual-view denoising strategy, employing low-rank SVD to the user-item interaction matrix for a denoised social graph and contrastive learning to align the original and reconstructed social graphs. Addressing the interest inconsistency between social and behavioral interests, we adopt a "mutual distillation" technique to isolate the original interests into aligned social/behavior interests and social/behavior specific interests, maximizing the utility of both. Experimental results on widely adopted industry datasets verify the method's effectiveness, particularly for cold users, offering a fresh perspective for future research. The implementation can be accessed at https://github.com/WANGLin0126/CLSRec.

Zhijian Feng, Wenhao Zheng, Xuanji Xiao

In real-world recommendation systems, users would engage in variety scenarios, such as homepages, search pages, and related recommendation pages. Each of these scenarios would reflect different aspects users focus on. However, the user interests may be inconsistent in different scenarios, due to differences in decision-making processes and preference expression. This variability complicates unified modeling, making multi-scenario learning a significant challenge. To address this, we propose a novel reinforcement learning approach that models user preferences across scenarios by modeling user interest evolution across multiple scenarios. Our method employs Double Q-learning to enhance next-item prediction accuracy and optimizes contrastive learning loss using Q-value to make model performance better. Experimental results demonstrate that our approach surpasses state-of-the-art methods in multi-scenario recommendation tasks. Our work offers a fresh perspective on multi-scenario modeling and highlights promising directions for future research.

Xuanji Xiao, Huabin Chen, Yuzhen Liu et al.

Click-through rate (CTR) and post-click conversion rate (CVR) predictions are two fundamental modules in industrial ranking systems such as recommender systems, advertising, and search engines. Since CVR involves much fewer samples than CTR (known as the CVR data sparsity problem), most of the existing works try to leverage CTR&CVR multi-task learning to improve CVR performance. However, typical coarse-grained sub-network/layer sharing methods may introduce conflicts and lead to performance degradation, since not every neuron or neuron connection in one layer should be shared between CVR and CTR tasks. This is because users may have different fine-grained content feature preferences between deep consumption and click behavior, represented by CVR and CTR, respectively. To address this sharing&conflict problem, we propose a novel multi-task CVR modeling scheme with neuron-connection level sharing named NCS4CVR, which can automatically and flexibly learn which neuron weights are shared or not shared without artificial experience. Compared with previous layer-level sharing methods, this is the first time that a fine-grained CTR&CVR sharing method at the neuron connection level is proposed, which is a research paradigm shift in the sharing level. Both offline and online experiments demonstrate that our method outperforms both the single-task model and the layer-level sharing model. Our proposed method has now been successfully deployed in an industry video recommender system serving major traffic.