Search

In recent years, millions of households have shifted from traditional TVs to smart TVs for viewing online videos on TV screens. In this article, we perform extensive analyses on a large‐scale online video viewing log on smart TVs. Because time influences almost every aspect of our lives, our aim is to understand temporal patterns of the online video viewing behavior of smart TV viewers at the crowd level. First, we measure the amount of time per hour spent in watching online videos on smart TV by each household on each day. By applying clustering techniques, we identify eight daily patterns whose peak hours occur in different segments of the day. The differences among households can be characterized by three types of temporal habits. We also uncover five periodic weekly patterns. There seems to be a circadian rhythm at the crow level. Further analysis confirms that there exists a holiday effect in the online video viewing behavior on smart TVs. Finally, we investigate the popularity variations of different video categories over the day. The obtained insights shed light on how we can partition a day to improve the performance of time‐aware video recommendations for smart TV viewers. [ABSTRACT FROM AUTHOR]

Most Web search diversity approaches can be categorized as Document Level Diversification ( DocLD ), Topic Level Diversification ( TopicLD ) or Term Level Diversification ( TermLD ). DocLD selects the relevant documents with minimal content overlap to each other. It does not take the coverage of query subtopics into account. TopicLD solves this by modeling query subtopics explicitly. However, the automatic mining of query subtopics is difficult. TermLD tries to cover as many query topic terms as possible, which reduces the task of finding a query's subtopics into finding a set of representative topic terms. In this paper, we propose a novel User Session Level Diversification ( UserLD ) approach based on the observation that a query's subtopics are implicitly reflected by the search intents in different user sessions. Our approach consists of two phases: (I) Session Graph Construction and (II) Diversity Reranking . For a given query, phase (I) builds a Session Graph which considers relevant user sessions and preliminary retrieval results as nodes and the nodes' pairwise similarities as edge weights. Phase (II) reranks the preliminary retrieval results by minimizing a Session Graph based diversity loss function. Extensive experiments on two standard datasets of NACSIS Test Collections for IR (NTCIR) demonstrate the effectiveness of our approach. The advantage of our approach lies in its ability of avoiding mining the query subtopics in advance while achieving almost the same or better performances compared with previous approaches. [ABSTRACT FROM AUTHOR]

Location information is very important in the real world. How to detect users' locations and further movement trajectories automatically is significant for many location-based services such as dietary recommendation and tourism planning. With the rapid development of social media such as Sina Weibo and Twitter, more and more people publish posts at any time which contain their real-time location information. This makes it possible to detect users' movement trajectories automatically by social media. In this paper, we propose a novel method to detect city-level locations and further identify movement trajectory for a given user in social media. Our efforts are made in two phases: (1) location detection, and (2) movement trajectory identification. Phase (1) combines two components, i.e. a Bayes model based on words distribution over locations and the social relationships of users in social media, to detect or infer users' locations. The former one is utilized to judge whether the content of a post contains explicit/implicit location information and further to detect the location. The latter one assumes that friends tend to gather together when they talk about the same thing. Thus, it infers a user's locations from his friends' detected locations in social media. Phase (2) constructs the user's movement trajectory by smoothing these detected locations in first phase according to both semantic context of posts and transfer time between locations. Experiments on real dataset from Sina Weibo demonstrate that our approach can outperform baseline methods significantly in terms of Precision, Recall, F1, Error Distance ( ErrDist), Detect Ratio ( DeteRatio), T-Precision, T-Recall and T-F1, respectively. [ABSTRACT FROM AUTHOR]

Query classification is an important part of exploring the characteristics of web queries. Existing studies are mainly based on Broder's classification scheme and classify user queries into navigational, informational, and transactional categories according to users' information needs. In this article, we present a novel classification scheme from the perspective of queries' temporal patterns. Queries' temporal patterns are inherent time series patterns of the search volumes of queries that reflect the evolution of the popularity of a query over time. By analyzing the temporal patterns of queries, search engines can more deeply understand the users' search intents and thus improve performance. Furthermore, we extract three groups of features based on the queries' search volume time series and use a support vector machine ( SVM) to automatically detect the temporal patterns of user queries. Extensive experiments on the Million Query Track data sets of the Text REtrieval Conference ( TREC) demonstrate the effectiveness of our approach. [ABSTRACT FROM AUTHOR]

How to understand intents behind user queries is crucial towards improving the performance of Web search systems. NTCIR-11 IMine task focuses on this problem. In this paper, we address the NTCIR-11 IMine task with two phases referred to as Query Intent Mining ( QIM) and Query Intent Ranking ( QIR). (I) QIM is intended to mine users' potential intents by clustering short text fragments related to the given query. (II) QIR focuses on ranking those mined intents in a proper way. Two challenges exist in handling these tasks. (II) How to precisely estimate the intent similarity between user queries which only consist of a few words. (2) How to properly rank intents in terms of multiple factors, e.g. relevance, diversity, intent drift and so on. For the first challenge, we first investigate two interesting phenomena by analyzing query logs and document datasets, namely ' Same-Intent-Co-Click' ( SICC) and ' Same-Intent-Similar-Rank' ( SISR). SICC means that when users issue different queries, these queries represent the same intent if they click on the same URL. SISR means that if two queries denote the same intent, we should get similar search results when issuing them to a search engine. Then, we propose similarity functions for QIM based on the two phenomena. For the second challenge, we propose a novel intent ranking model which considers multiple factors as a whole. We perform extensive experiments and an interesting case study on the Chinese dataset of NTCIR-11 IMine task. Experimental results demonstrate the effectiveness of our proposed approaches in terms of both QIM and QIR. [ABSTRACT FROM AUTHOR]

Abstract: Focused crawling is increasingly seen as a solution to address the scalability limitations of existing general-purpose search engines, by traversing the Web to only gather pages that are relevant to a specific topic. How to predict the relevance of the unvisited pages pointed to by candidate URLs in the crawling frontier to a given topic is a key issue in the design of focused crawlers. In this paper, we propose a novel approach based on multiple relevance prediction strategies to address this problem. For cross-language crawling, we first introduce a hierarchical taxonomy to describe topics in both English and Chinese. We then present a formal description of the relevance predicting process and discuss four strategies that make use of page contents, anchor texts, URL addresses and link types of Web pages, respectively, to evaluate the relevance more accurately, in which we propose a particular strategy using Chinese URL addresses to estimate the relevance of cross-language Web pages. Finally, we get a new focused crawling algorithm (FCMRPS, Focused Crawling based on Multiple Relevance Prediction Strategies) based on the combination of these strategies and Shark-Search, which is a classic focused crawling algorithm. Experiments show that the FCMRPS is more effective than the traditional algorithms, namely Breadth-First, Best-First and Shark-Search, in terms of precision and sum of information. [Copyright &y& Elsevier]

The article focuses on addressing the challenge of enhancing recommendation diversity in sequential recommendation (SR) models, particularly for short interaction sequences where limited information hampers accurate modeling of user intents. It mentions the proposed TRIER model tackles this by retrospectively predicting historical interactions to augment sequences and prospectively capturing potential intents.

With the recent prevalence of online fashion-oriented communities and advances in multimedia processing, increasing research interests have been paid to the fashion compatibility modeling, where the compatibility between complementary fashion items (e.g., a top and a bottom) can be assessed automatically. Existing fashion compatibility modeling techniques mainly focus on measuring the compatible preference between fashion items with Deep Neural Networks (DNN), but overlook the generative compatibility modeling. Differently, in this paper, we explore the potential of the Generative Adversarial Network (GAN) in fashion compatibility modeling and thus propose a Multi-modal Generative Compatibility Modeling (MGCM) scheme. In particular, we introduce a multi-modal enhanced compatible template generation network, regularized by the pixel-wise consistency and template compatibility, to sketch a compatible template as the auxiliary link between fashion items. Accordingly, MGCM is able to measure the compatibility between complementary fashion items comprehensively from both item-item and item-template perspectives. Experimental results on two real-world datasets demonstrate the superiority of the proposed scheme over state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Most previous work on outfit recommendation focuses on designing visual features to enhance recommendations. Existing work neglects user comments of fashion items, which have been proven to be effective in generating explanations along with better recommendation results. We propose a novel neural network framework, neural outfit recommendation (NOR), that simultaneously provides outfit recommendations and generates abstractive comments. Neural outfit recommendation (NOR) consists of two parts: outfit matching and comment generation. For outfit matching, we propose a convolutional neural network with a mutual attention mechanism to extract visual features. The visual features are then decoded into a rating score for the matching prediction. For abstractive comment generation, we propose a gated recurrent neural network with a cross-modality attention mechanism to transform visual features into a concise sentence. The two parts are jointly trained based on a multi-task learning framework in an end-to-end back-propagation paradigm. Extensive experiments conducted on an existing dataset and a collected real-world dataset show NOR achieves significant improvements over state-of-the-art baselines for outfit recommendation. Meanwhile, our generated comments achieve impressive ROUGE and BLEU scores in comparison to human-written comments. The generated comments can be regarded as explanations for the recommendation results. We release the dataset and code to facilitate future research. [ABSTRACT FROM AUTHOR]

Clothing has gradually become the beauty enhancing product, while the harmonious clothing matching is critical for a suitable outfit. The existing clothing matching techniques mainly rely on the visual features but overlook the textual metadata, which may be insufficient to comprehensively encode the fashion items. Nowadays, fashion experts are enabled to share their fashion tips by demonstrating their outfit compositions on the fashion-oriented online communities. Each outfit usually consists of several complementary fashion items (e.g., a top, a bottom and a pair of shoes), which involves an image along with the textual metadata (e.g., the categories and titles). The rich fashion data provide us an opportunity for the clothing matching, especially the complementary fashion item matching. In this work, we propose a multiple autoencoder neural network based on the Bayesian Personalized Ranking, dubbed BPR-MAE. Seamlessly exploring the multi-modalities (i.e., the visual and textual modalities) of fashion items, this framework is able to not only comprehensively model the compatibility between fashion items (e.g., tops and bottoms, bottoms and shoes) but also fulfill the complementary fashion item matching among multiple fashion items. Experimental results on the real-world dataset FashionVC+ demonstrate the effectiveness of BPR-MAE, based on which we provide certain deep insights that can benefit the future research. [ABSTRACT FROM AUTHOR]

Recommender systems with social networks have been well studied in recent years. However, most of these methods ignore the social contextual information among users and items, which is significant and useful for predicting users' preferences in many recommendation problems. Moreover, most existing social recommendation methods have been proposed for the scenarios where users can provide explicit ratings. But in fact, the explicit feedback is not always available, most of the feedback in real social networks is not explicit but implicit. Motivated by above observations, we propose a unified ranking framework fusing social contextual information and common social relations for implicit feedback. Specifically, we first extend the user latent features by the implicit interest deduced from social context, and then we integrate the common social relations as factorization terms to further improve recommendation quality. Finally, we optimize our model in a Bayesian personalized ranking framework. The experiments on real-world dataset show that our approach outperforms the other state-of-the-art algorithms in terms of AUC, NDCG and Pre@3. This result demonstrates the importance of social context and common social relations for the formation of the implicit ratings. [ABSTRACT FROM AUTHOR]

Abstract: Ranking nodes by their spreading ability in complex networks is a fundamental problem which relates to wide applications. Local metric like degree centrality is simple but less effective. Global metrics such as betweenness and closeness centrality perform well in ranking nodes, but are of high computational complexity. Recently, to rank nodes effectively and efficiently, a semi-local centrality measure has been proposed as a tradeoff between local and global metrics. However, in semi-local centrality, only the number of the nearest and the next nearest neighbors of a node is taken into account, while the topological connections among the neighbors are neglected. In this paper, we propose a local structural centrality measure which considers both the number and the topological connections of the neighbors of a node. To evaluate the performance of our method, we use the Susceptible–Infected–Recovered (SIR) model to simulate the epidemic spreading process on both artificial and real networks. By measuring the rank correlation between the ranked list generated by simulation results and the ones generated by centrality measures, we show that our method can rank the spreading ability of nodes more accurately than centrality measures such as degree, -shell, betweenness, closeness and local centrality. Further, we show that our method can better distinguish the spreading ability of nodes. [Copyright &y& Elsevier]

The wheeled robots have been successfully applied in many aspects, such as industrial handling vehicles, and wheeled service robots. To improve the safety and reliability of wheeled robots, this paper presents a novel hybrid fault diagnosis framework based on Mittag-Leffler kernel (ML-kernel) support vector machine (SVM) and Dempster-Shafer (D-S) fusion. Using sensor data sampled under different running conditions, the proposed approach initially establishes multiple principal component analysis (PCA) models for fault feature extraction. The fault feature vectors are then applied to train the probabilistic SVM (PSVM) classifiers that arrive at a preliminary fault diagnosis. To improve the accuracy of preliminary results, a novel ML-kernel based PSVM classifier is proposed in this paper, and the positive definiteness of the ML-kernel is proved as well. The basic probability assignments (BPAs) are defined based on the preliminary fault diagnosis results and their confidence values. Eventually, the final fault diagnosis result is archived by the fusion of the BPAs. Experimental results show that the proposed framework not only is capable of detecting and identifying the faults in the robot driving system, but also has better performance in stability and diagnosis accuracy compared with the traditional methods. [ABSTRACT FROM AUTHOR]