Yan Yang

An Empirical Study of Personal Factors and Social Effects on Rating Prediction

In social networks, the link between a pair of friends has been reported effective in improving recommendation accuracy. Previous studies mainly based on the assumption that any pair of friends shall have similar interests, via minimizing the gap between user’s taste and the average (or similar) taste of this user’s friends to reduce the error of rating prediction. However, these methods ignore the diversity of user’s taste. In this paper, we focus on learning the diversity of user’s taste and effects from this user’s friends in terms of rating behavior. We propose a novel recommendation approach, namely Personal factors with Weighted Social effects Matrix Factorization (PWS), which utilities both user’s taste and social effects to provide recommendations. Experimental results carried out on 3 datasets, show the effectiveness of the proposed approach.

User Identification within a Shared Account: Improving IP-TV Recommender Performance

Multiple users share a common account in Internet Protocol Television (IP-TV) services. Can such shared accounts be identified solely on the basis of logs recorded by set top boxes (STBs)? Once a shared account is identified, can the different users sharing it be identified as well? We suppose different users within a shared account not only have different preferences for TV programs, but also get used to consuming services in different periods (e.g., after dinner or at weekend). We propose an algorithm to decompose users in composite accounts based on mining different preferences over different periods from consumption logs. In our experiments, the proposed algorithm outperforms traditional user-based collaborative filtering method 3-8 times when leveraging the decomposed users for personalized recommendation.

Adaptive Temporal Model for IPTV Recommendation

How to help the IPTV service provider make the program recommendation to their clients is the problem we propose to solve in this paper. Here we offer an adaptive temporal model to identify multiple members under a shared IPTV account. The time intervals are first detected and defined in each account. Then, the preference similarity is calculated among the intervals to extract the members. After that, we evaluate our model on the industrial data sets by a famous IPTV provider. The experimental results show that our proposed model is promising and outperform the state-of-the-art algorithms with low computational complexity and versatility without user feedback. Furthermore, the proposed model has been officially adopted by the IPTV provider and applied in their IPTV systems with excellent user satisfaction in 2013.

Answering Range-Based Reverse kNN Queries.

Given a point q, a reverse k nearest neighbor (RkNN) query retrieves all the data points that have q as one of their k nearest neighbors. Despite significant progress on this problem, there is a research gap in finding RkNNs not just for an object, but for a given range, which is a natural extension of the problem. Motivated by this, we develop algorithms for exact processing of range-based RkNN with arbitrary values of k on dynamic datasets, which retrieve all the data points that have any position in the given query range R as one of their k nearest neighbors. The experimental results demonstrate the efficiency and the accuracy of our proposed optimizations and algorithms.

Cleaning uncertain graphs via noisy crowdsourcing

Uncertain graph is an important data model for many real-world applications. To answer the query on the uncertain graphs, the edges in these graphs are associated with existential probabilities that represent the likelihood of the existence of the edge. Almost all works on this area focus on how to promote the efficiency of the query processing. However, another issue should be notable, that is, the query results from the uncertain graphs are sometimes uninformative due to the edge uncertainty. We adopt a crowdsourcing-based approach to make the query results more informative. To save the monetary and time cost of crowdsourcing, we should select the optimal edges to clean to maximize the quality improvement. However, the noise of the crowdsourcing results will make the problem more complex. We prove that the problem is #P-hard and propose an efficient algorithm to derive the optimal edge. Our experimental results show that our proposed algorithm outperforms random-selection up to 22 times in quality improvement and each-edge-comparison way up to 5 times fast in elapsed time, which proves this algorithm is both effective and efficient.

Representation Learning with Entity Topics for Knowledge Graphs

Knowledge representation learning which represents triples as semantic embeddings has achieved tremendous success these years. Recent work aims at integrating the information of triples with texts, which has shown great advantages in alleviating the data sparsity problem. However, most of these methods are based on word-level information such as co-occurrence in texts, while ignoring the latent semantics of entities. In this paper, we propose an entity topic based representation learning (ETRL) method, which enhances the triple representations with the entity topics learned by the topic model. We evaluate our proposed method knowledge graph completion task. The experimental results show that our method outperforms most state-of-the-art methods. Specifically, we achieve a maximum improvement of 7.9% in terms of hits@10.

Knowledge Memory Based LSTM Model for Answer Selection

Recurrent neural networks (RNN) have shown great success in answer selection task in recent years. Although the attention mechanism has been widely used to enhance the information interaction between questions and answers, knowledge is still the gap between their representations. In this paper, we propose a knowledge memory based RNN model, which incorporates the knowledge learned from the data sets into the question representations. Experiments on two benchmark data sets show the great advantages of our proposed model over that without the knowledge memory. Furthermore, our model outperforms most of the recent progress in question answering.