Xiaolin Zheng

A hybrid approach for movie recommendation via tags and ratings

A hybrid recommendation approach for movies via tags and ratings was proposed.Social tags were reconstructed according to user preference based on content annotation.Our model improved the ability of fusion by applying the potential aspects.Our hybrid method significantly outperforms comparative recommendation approaches. Selecting a movie often requires users to perform numerous operations when faced with vast resources from online movie platforms. Personalized recommendation services can effectively solve this problem by using annotating information from users. However, such current services are less accurate than expected because of their lack of comprehensive consideration for annotation. Thus, in this study, we propose a hybrid movie recommendation approach using tags and ratings. We built this model through the following processes. First, we constructed social movie networks and a preference-topic model. Then, we extracted, normalized, and reconditioned the social tags according to user preference based on social content annotation. Finally, we enhanced the recommendation model by using supplementary information based on user historical ratings. This model aims to improve fusion ability by applying the potential effect of two aspects generated by users. One aspect is the personalized scoring system and the singular value decomposition algorithm, the other aspect is the tag annotation system and topic model. Experimental results show that the proposed method significantly outperforms three categories of recommendation approaches, namely, user-based collaborative filtering (CF), model-based CF, and topic model based CF.

URL Rule Based Focused Crawler

Vertical search engines use focused crawlers as their key component and develop some specific algorithms to select Web pages relevant to some pre-defined set of topics. Therefore, how to effectively build up a semantic pattern for specific topics is extremely important to such search engines. In this paper we propose an UBFC (URL rule based focused crawler) algorithm based on a double-crawler framework (an experimental crawler and a focused crawler). The kernel of our UBFC is an URL regular expression learner, which is used to automatically learn and generalize the regular expressions of URLs of the sample Web pages. The so generated URL regular expressions are used to direct the UBFC to work. Using UBFC, we developed a Nutch based focused crawler for hardwaretoday.com and get an excellent result. To evaluate the UBFC, we conduct some experiments to make an analysis through comparison with two proposed methods, the BFSC (breadth first search crawler) and BLFC (baseline focused crawler).

Recommender System Based on Social Trust Relationships

The development of social network has increased the importance of social recommendation. However, social recommender systems have only recently been given research attention. Social relationships between users, especially trust relationships, can facilitate the design of social recommender systems. Such systems are based on the idea that users linked by a social network tend to share similar interests. Existing recommender approaches based on social trust relationships do not fully utilize such relationships and thus have low prediction accuracy or slow convergence speed. We propose a factor analysis approach that explicitly and implicitly uses social trust relationships simultaneously to overcome this limitation and fully utilize social trust relationships. Our approach combines the advantages of the existing two approaches, social recommendation using probabilistic matrix factorization and learning to recommend with social trust ensemble. Based on Epinions data sets, our approach has both significantly higher prediction accuracy and convergence speed than traditional collaborative filtering technology and state-of-the-art trust-based recommendation approaches.

Policy-Based Web Service Selection in Context Sensitive Environment

Web service selection is a key issue in the service-oriented computing. Generally, Web services are provided with different QoS values, so they can be selected dynamically in service composition process. However, the conventional context free QoS selection model does not consider the context sensitive constraints and the changeability of QoS values. This paper proposes a policy-based context sensitive QoS model to effectively calculate QoS values and select web services. By introducing context to the QoS model, web service composition can be used widely and flexibly in the real world business.

Latent space regularization for recommender systems

The primary latent factor model cannot effectively optimize the user-item latent spaces because of the sparsity and imbalance of the rating data. Although existing studies have focused on exploring auxiliary information for users or items, few researchers have considered users and items jointly. For instance, social information is incorporated into models without considering the item side. In this paper, we introduce latent space regularization (LSR) and provide a general method to improve recommender systems (RSs) by incorporating LSR. We take the assumption that users prefer items that cover one or several topics that they are interested in, instead of all the topics, which reflects real-world situations. For instance, a user may focus on the humorous part of an item when he or she is at leisure time, regardless of the relevance of the item to his research topics. LSR operates from this assumption to account for both the user and item sides simultaneously. From another point of view, LSR is likely to improve the Tanimoto similarity of observed user-item pairs. As a result, LSR utilizes the number of ratings in a manner similar to weighted matrix factorization. We incorporate LSR into both the traditional collaborative filtering models that use only rating information and the collaborative filtering model that uses auxiliary content information as two examples. Experimental results from on two real-world datasets show not only the superiority of our model over other regularization models, but also its effectiveness and the possibility of incorporating it into various existing latent factor models.

Topic tensor factorization for recommender system

Reviews are collaboratively generated by users on items and generally contain rich information than ratings in a recommender system scenario. Ratings are modeled successfully with latent space models by capturing interaction between users and items. However, only a few models collaboratively deal with documents such as reviews. In this study, by modeling reviews as a three-order tensor, we propose a refined tensor topic model (TTM) for text tensors inspired by Tucker decomposition. User and item dimensions are co-reduced with vocabulary space, and interactions between users and items are captured using a core tensor in dimension-reduced form. TTM is proposed to obtain low-rank representations of words as well as of users and items. Furthermore, general rules are developed to transform a decomposition model into a probabilistic model. TTM is augmented further to predict ratings with the assistance of a low-dimensional representation of users and items obtained by TTM. This augmented model is called matrix factorization by learning a bilinear map. A core regularized version is further developed to incorporate additional information from the TTM. Encouraging experimental results not only show that the TTM outperforms existing topic models in modeling texts with a user-item-word structure, but also show that our proposed rating prediction models outperform state-of-the-art approaches.

A Hybrid Trust-Based Recommender System for Online Communities of Practice

The needs for life-long learning and the rapid development of information technologies promote the development of various types of online Community of Practices (CoPs). In online CoPs, bounded rationality and metacognition are two major issues, especially when learners face information overload and there is no knowledge authority within the learning environment. This study proposes a hybrid, trust-based recommender system to mitigate above learning issues in online CoPs. A case study was conducted using Stack Overflow data to test the recommender system. Important findings include: (1) comparing with other social community platforms, learners in online CoPs have stronger social relations and tend to interact with a smaller group of people only; (2) the hybrid algorithm can provide more accurate recommendations than celebrity-based and content-based algorithm and; (3) the proposed recommender system can facilitate the formation of personalized learning communities.

Trust Based Service Selection in Service Oriented Environment

In a service-oriented environment, it is inevitable and indeed quite common to deal with web services, whose reliability is unknown to the users. The reputation system is a popular technique currently used for providing a global quality score of a service provider to requesters. However, such global information is far from sufficient for service requesters to choose the most qualified services. In order to tackle this problem, the authors present a trust based architecture containing a computational trust model for quantifying and comparing the trustworthiness of services. In this trust model, they firstly construct a network based on the direct trust relations between participants and rating similarity in service oriented environments, then propose an algorithm for propagating trust in the social network based environment which can produce personalized trust information for a specific service requester, and finally implement the trust model and simulate various malicious behaviors in not only dense but also sparse networks which can verify the attack-resistant and robustness of the proposed approach. The experiment results also demonstrate the feasibility and benefit of the approach.

A Recommender System Model Combining Trust with Topic Maps

Recommender Systems (RS) aim to suggest users with items that they might like based on users’ opinion on items. In practice, information about the users’ opinion on items is usually sparse compared to the vast information about users and items. Therefore it is hard to analyze and justify users’ favorites, particularly those of cold start users. In this paper, we propose a trust model based on the user trust network, which is composed of the trust relationships among users. We also introduce the widely used conceptual model Topic Map, with which we try to classify items into topics for Recommender analysis. We novelly combine trust relations among users with Topic Maps to resolve the sparsity problem and cold start problem. The evaluation shows our model and method can achieve a good recommendation effect.

An Ontology-Driven Discovery Architecture to Support Service Composition

In the service collaboration field, there exists the need to combine two or more services together to fulfill a complex goal. Though service composition is an active research field, most of solutions are focused on Web services and lack of the consideration for business-level collaboration. In this paper we present a discovery architecture that supports business-level service composition. In our architecture, ontologies are adopted to give well-defined semantic meanings for knowledge, service offers and requests. According to decomposition templates and ontologies, the service goal of a request is decomposed to several sub-goals. Then the services that match sub-goals can be discovered in isolation. Finally, the semantic compatibility between relevant services is checked and the feasible service combinations that fulfill the whole request can be formed. We discuss the discovery approach, and show the possibility of fulfilling a request with the composition of services.