Reviews information is dominant for users to make online purchasing decisions in e-commerces. However, the usefulness of reviews varied. We argue that less-useful hurt model's performance, and are also less meaningful user's reference. While some existing models utilize improving performance recommender systems, few them consider recommendation quality. In this paper, we introduce a novel attention mechanism explore reviews, propose Neural Attentional Regression model with Review-level...

Significance Scientific peer review has been a cornerstone of the scientific method since 1600s. Debate continues regarding merits single-blind review, in which anonymous reviewers know authors paper and their affiliations, compared with double-blind this information is hidden. We present an experimental study question. In computer science, research often appears first or exclusively peer-reviewed conferences rather than journals. Our considers full-length submissions to highly selective...

Recommender Systems have become a very useful tool for large variety of domains. Researchers been attempting to improve their algorithms in order issue better predictions the users. However, one current challenges area refers how properly evaluate generated by recommender system. In extent offline evaluations, some traditional concepts evaluation explored, such as accuracy, Root Mean Square Error and P@N top-k recommendations. recent years, more research proposed new novelty, diversity...

Explainability and effectiveness are two key aspects for building recommender systems. Prior efforts mostly focus on incorporating side information to achieve better recommendation performance. However, these methods have some weaknesses: (1) prediction of neural network-based embedding hard explain debug; (2) symbolic, graph-based approaches (e.g., meta path-based models) require manual domain knowledge define patterns rules, ignore the item association types (e.g. substitutable...

Recommendation systems play a vital role to keep users engaged with personalized contents in modern online platforms. Recently, deep learning has revolutionized many research fields and there is surge of interest applying it for recommendation. However, existing studies have largely focused on exploring complex deep-learning architectures recommendation task, while typically the negative sampling strategy model learning. Despite effectiveness, we argue that these methods suffer from two...

Recommender systems are an essential tool to relieve the information overload challenge and play important role in people's daily lives. Since recommendations involve allocations of social resources (e.g., job recommendation), issue is whether fair. Unfair not only unethical but also harm long-term interests recommender system itself. As a result, fairness issues have recently attracted increasing attention. However, due multiple complex resource allocation processes various definitions,...

Traditional studies on recommender systems usually leverage only one type of user behaviors (the optimization target, such as purchase), despite the fact that users also generate a large number various types interaction data (e.g., view, click, add-to-cart, etc). Generally, these heterogeneous multi-relational provide well-structured information and can be used for high-quality recommendation. Early efforts towards leveraging fail to capture high-hop structure user-item interactions, which...

Many previous studies attempt to utilize information from other domains achieve better performance of recommendation. Recently, social has been shown effective in improving recommendation results with transfer learning frameworks, and the part helps learn users' preferences both item domain domain. However, two vital issues have not well-considered existing methods: 1) Usually, a static scheme is adopted share user's common preference between domains, which robust real life where degrees...

Recent studies on recommendation have largely focused exploring state-of-the-art neural networks to improve the expressiveness of models, while typically apply Negative Sampling (NS) strategy for efficient learning. Despite effectiveness, two important issues not been well-considered in existing methods: 1) NS suffers from dramatic fluctuation, making sampling-based methods difficult achieve optimal ranking performance practical applications; 2) although heterogeneous feedback (e.g., view,...

Social connections are known to be helpful for modeling users' potential preferences and improving the performance of recommender systems. However, in social-aware recommendations, there two issues which influence inference preferences, haven't been well-studied most existing methods: First, a user may only partially match that his friends certain aspects, especially when considering with diverse interests. Second, an individual, strength might different, as not all equally system. To...

Traditional recommender systems mainly aim to model inherent and long-term user preference, while dynamic demands are also of great importance. Typically, a historical consumption will have impacts on the for its relational items. For instance, users tend buy complementary items together (iPhone Airpods) but not substitutive (Powerbeats Airpods), although substitutes bought one still cater his/her preference. To better effects history sequence, previous studies introduce semantics item...

Knowledge tracing (KT) aims to model students' knowledge level based on their historical performance, which plays an important role in computer-assisted education and adaptive learning. Recent studies try take temporal effects of past interactions into consideration, such as the forgetting behavior. However, existing work mainly relies time-related features or a global decay function time-sensitive effects. Fine-grained dynamics different cross-skill impacts have not been well studied (named...

Legal case retrieval is of vital importance for ensuring justice in different kinds law systems and has recently received increasing attention information (IR) research. However, the relevance judgment criteria previous datasets are either not applicable to non-cited relationship cases or instructive enough future follow. Besides, most existing benchmark do focus on selection queries. In this paper, we construct Chinese Case Retrieval Dataset (LeCaRD), which contains 107 query over 43,000...

Recently, Information Retrieval community has witnessed fast-paced advances in Dense (DR), which performs first-stage retrieval with embedding-based search. Despite the impressive ranking performance, previous studies usually adopt brute-force search to acquire candidates, is prohibitive practical Web scenarios due its tremendous memory usage and time cost. To overcome these problems, vector compression methods have been adopted many applications. One of most popular Product Quantization...

Recommender systems provide essential web services by learning users' personal preferences from collected data. However, in many cases, also need to forget some training From the perspective of privacy, users desire a tool erase impacts their sensitive data trained models. utility, if system's utility is damaged bad data, system needs such regain utility. While unlearning very important, it has not been well-considered existing recommender systems. Although there are researches have studied...

In modern search engines, an increasing number of result pages (SERPs) are federated from multiple specialized engines (called verticals, such as Image or Video). As effective approach to interpret users' click-through behavior feedback information, most click models were designed reduce the position bias and improve ranking performance ordinary results, which have homogeneous appearances. However, when vertical results combined with ones, significant differences in presentation may lead...

Relevance is a fundamental concept in information retrieval (IR) studies. It however often observed that relevance as annotated by secondary assessors may not necessarily mean usefulness and satisfaction perceived users. In this study, we confirm the difference laboratory study which collect annotations external assessors, user users, for set of search tasks. We also find measure based on rather than has better correlation with satisfaction. However, show are capable annotating when provided...

The frequently changing user preferences and/or item profiles have put essential importance on the dynamic modeling of users and items in personalized recommender systems. However, due to insufficiency per user/item records when splitting already sparse data across time dimension, previous methods restrict drifting purchasing patterns pre-assumed distributions, were hardly able model them rather directly with, for example, series analysis. Integrating content information helps alleviate...

Satisfaction prediction is one of the prime concerns in search performance evaluation. It a non-trivial task for two major reasons: (1) The definition satisfaction rather subjective and different users may have opinions judgement. (2) Most existing studies on mainly rely users' click-through or query reformulation behaviors but there are many sessions without such kind interactions. To shed light these research questions, we construct an experimental engine that could collect feedback as...

Repeat consumption is a common scenario in daily life, such as repurchasing items and revisiting websites, critical factor to be taken into consideration for recommender systems. Temporal dynamics play important roles modeling repeat consumption. It noteworthy that with distinct lifetimes, consuming tendency the next one fluctuates differently time. For example, users may repurchase milk weekly, but it possible mobile phone after long period of Therefore, how adaptively incorporate various...

As online shopping becomes increasingly popular, users perform more product search to purchase items. Previous studies have investigated people's behaviours and ways predict purchases. However, from a user perspective, there still lacks an in-depth understanding of why search, how they interact with, perceive the results. In this paper, we conduct both study log analysis address following three questions: (1) what are intents underlying their activities? (2) do behave differently under...

As queries submitted by users directly affect search experiences, how to organize has always been a research focus in Web studies. While request becomes complex and exploratory, many sessions contain more than single query thus reformulation necessity. To help better formulate their these tasks, modern engines usually provide series of entries on engine result pages (SERPs), i.e., suggestions related entities. However, few existing work have thoroughly studied why perform reformulations...