Precise user and item embedding learning is the key to building a successful recommender system. Traditionally, Collaborative Filtering (CF) provides way learn embeddings from user-item interaction history. However, performance limited due sparseness of behavior data. With emergence online social networks, systems have been proposed utilize each user's local neighbors' preferences alleviate data sparsity for better modeling. We argue that, platform, her potential influenced by trusted users,...

Social recommendation has emerged to leverage social connections among users for predicting users' unknown preferences, which could alleviate the data sparsity issue in collaborative filtering based recommendation. Early approaches relied on utilizing each user's first-order neighbors' interests better user modeling, and failed model influence diffusion process from global network structure. Recently, we propose a preliminary work of neural <italic...

Most modern recommender systems predict users preferences with two components: user and item embedding learning, followed by the user-item interaction modeling. By utilizing auxiliary review information accompanied ratings, many of existing review-based recommendation models enriched user/item learning ability historical reviews or better modeled interactions help available target reviews. Though significant progress has been made, we argue that current solutions for suffer from drawbacks....

Collaborative filtering(CF) is one of the most popular techniques for building recommender systems. To alleviate data sparsity issue in CF, social recommendation has emerged by leveraging influence among users better performance. In these systems, users' preferences over time are determined their temporal dynamic interests as well general static interests. meantime, complex interplay between internal and from network drives evolution time. Nevertheless, traditional approaches either...

Collaborative filtering (CF) is one of the most popular techniques for building recommender systems. To overcome data sparsity in CF, social systems have emerged to boost recommendation performance by utilizing correlation among users' interests. Recently, inspired immense success deep learning embedding learning, neural network-based shown promising performance. Nevertheless, few researchers attempted tackle problem with models. this end, paper, we design a architecture that organically...

In many recommender systems, users express item opinions through two kinds of behaviors: giving preferences and writing detailed reviews. As both behaviors reflect users' assessment items, review enhanced systems leverage these user to boost recommendation performance. On the one hand, researchers proposed better model embeddings with additional information for enhancing preference prediction accuracy. other some recent works focused on automatically generating reviews explanations related...

Multimedia-based recommendation is a challenging task that requires not only learning collaborative signals from user-item interaction, but also capturing modality-specific user interest clues complex multimedia content. Though significant progress on this challenge has been made, we argue current solutions remain limited by multimodal noise contamination. Specifically, considerable proportion of content irrelevant to the preference, such as background, overall layout, and brightness images;...

Collaborative Filtering (CF) is one of the most successful approaches for recommender systems. With emergence online social networks, recommendation has become a popular research direction. Most these models utilized each user's local neighbors' preferences to alleviate data sparsity issue in CF. However, they only considered neighbors user and neglected process that users' are influenced as information diffuses network. Recently, Graph Convolutional Networks~(GCN) have shown promising...

While effective in recommendation tasks, collaborative filtering (CF) techniques face the challenge of data sparsity. Researchers have begun leveraging contrastive learning to introduce additional self-supervised signals address this. However, this approach often unintentionally distances target user/item from their neighbors, limiting its efficacy. In response, we propose a solution that treats neighbors anchor node as positive samples within final objective loss function. This paper...

In the rapidly evolving landscape of large language models (LLMs), most research has primarily viewed them as independent individuals, focusing on assessing their capabilities through standardized benchmarks and enhancing general intelligence. This perspective, however, tends to overlook vital role LLMs user-centric services in human-AI collaboration. gap becomes increasingly critical become more integrated into people's everyday professional interactions. study addresses important need...

With the surge in mobile gaming, accurately predicting user spending on newly downloaded games has become paramount for maximizing revenue. However, inherently unpredictable nature of behavior poses significant challenges this endeavor. To address this, we propose a robust model training and evaluation framework aimed at standardizing data to mitigate label variance extremes, ensuring stability modeling process. Within framework, introduce collaborative-enhanced designed predict game without...

Review information has been demonstrated beneficial for the explainable recommendation. It can be treated as training corpora generation-based methods or knowledge bases extraction-based models. However, methods, sparsity of user-generated reviews and high complexity generative language models lead to a lack personalization adaptability. For focusing only on relevant attributes makes them invalid in situations where explicit attribute words are absent, limiting potential

News recommendation aims to help online news platform users find their preferred articles. Existing methods usually learn models from historical user behaviors on news. However, these are biased providers. Models trained data may capture and even amplify the biases providers, unfair for some minority In this paper, we propose a provider fairness-aware framework (named ProFairRec), which can fair different providers data. The core idea of ProFairRec is provider-fair representations achieve...

Collaborative Filtering~(CF) typically suffers from the significant challenge of popularity bias due to uneven distribution items in real-world datasets. This leads a accuracy gap between popular and unpopular items. It not only hinders accurate user preference understanding but also exacerbates Matthew effect recommendation systems. To alleviate bias, existing efforts focus on emphasizing or separating correlation item representations their popularity. Despite effectiveness, works still...

As its availability and generality in online services, implicit feedback is more commonly used recommender systems. However, usually presents noisy samples real-world recommendation scenarios (such as misclicks or non-preferential behaviors), which will affect precise user preference learning. To overcome the problem, a popular solution based on dropping model training phase, follows observation that have higher losses than clean samples. Despite effectiveness, we argue this still has...

Social recommendation has emerged to leverage social connections among users for predicting users' unknown preferences, which could alleviate the data sparsity issue in collaborative filtering based recommendation. Early approaches relied on utilizing each user's first-order neighbors' interests better user modeling and failed model influence diffusion process from global network structure. Recently, we propose a preliminary work of neural (i.e., DiffNet) (Diffnet), models recursive capture...

Ranking ensemble is a critical component in real recommender systems. When user visits platform, the system will prepare several item lists, each of which generally from single behavior objective recommendation model. As multiple intents, e.g., both clicking and buying some specific category, are commonly concurrent visit, it necessary to integrate single-objective ranking lists into one. However, previous work on rank aggregation mainly focused fusing homogeneous with same while ignoring...

The explainability of recommender systems has attracted significant attention in academia and industry. Many efforts have been made for explainable recommendations, yet evaluating the quality explanations remains a challenging unresolved issue. In recent years, leveraging LLMs as evaluators presents promising avenue Natural Language Processing tasks (e.g., sentiment classification, information extraction), they perform strong capabilities instruction following common-sense reasoning....

Review based recommendation utilizes both users’ rating records and the associated reviews for recommendation. Recently, with rapid demand explanations of results, are used to train encoder–decoder models explanation text generation. As most general without detailed evaluation, some researchers leveraged auxiliary information users or items enrich generated text. Nevertheless, data is not available in scenarios may suffer from privacy problems. In this article, we argue that contain abundant...

Short video~(SV) online streaming has been one of the most popular Internet applications in recent years. When browsing SVs, users gradually immerse themselves and derive relaxation or knowledge. Whereas prolonged will lead to a decline positive feelings, continue due inertia, resulting decreased satisfaction. Immersion is shown be an essential factor for users' experience highly related interactions film, games, virtual reality. However, immersion SV interaction still unexplored, which...