We consider the problem of personalized text-based music retrieval where users' history of preferences are taken into account in addition to their issued textual queries.Current retrieval methods mostly rely on songs meta-data. This limits the query vocabulary. Moreover, it is very costly to gather this information in large collections of music. Alternatively, we use music annotations retrieved from social tagging Websites such as last.fm and use them as textual descriptions of songs. Considering a user's profile and using preference patterns of music among all users, as in collaborative filtering approaches, can be useful in providing personalized and more satisfactory results. The main challenge is how to include both users' profiles and the songs meta-data in the retrieval model. In this paper, we propose a hierarchical probabilistic model that takes into account the users' preference history as well as tag co-occurrences in songs. Our model is an extension of LDA where topics are formed as joint clusterings of songs and tags. These topics capture the tag associations and user preferences and correspond to different music tastes. Each user's profile is represented as a distribution over topics which shows the user's interests in different types of music.We will explain how our model can be used for contextual retrieval. Our experimental results show significant improvement in retrieval when user profiles are taken into account.

Traditional recommender systems try to provide users with recommendations which maximize the probability that the user will accept them. Recent studies have shown that recommender systems have a positive effect on the provider’s revenue. In this paper we show that by giving a different set of recommendations, the recommendation system can further increase the business’ utility (e.g. revenue), without any significant drop in user satisfaction. Indeed, the recommendation system designer should have in mind both the user, whose taste we need to reveal, and the business, which wants to promote specific content. In order to study these questions, we performed a large body of experiments on Amazon Mechanical Turk. In each of the experiments, we compare a commercial state-of-the-art recommendation engine with a modified recommendation list, which takes into account the utility (or revenue) which the business obtains from each suggestion that is accepted by the user. We show that the modified recommendation list is more desirable for the business, as the end result gives the business a higher utility (or revenue). To study possible longterm effects of giving the user worse suggestions, we asked the users how they perceive the list of recommendation that they received. Our findings are that any difference in user satisfaction between the list is negligible, and not statistically significant. We also uncover a phenomenon where movie consumers prefer watching and even paying for movies that they have already seen in the past than movies that are new to them.

In pervasive environments, users are situated in rich context and can interact with their surroundings through various services. To improve user experience in such environments, it is essential to find the services that satisfies user preferences in certain context. Thus the suitability of discovered services is highly dependent on how much the context-aware system can understand users' current context and preferred activities. In this paper, we propose an unsupervised learning solution for mining user preferences from the user's past context. To cope with the high dimensionality and heterogeneity of context data, we propose a subspace clustering approach that is able to find user preferences identified by different feature sets. The results of our approach are validated by a series of experiments.

In this paper, we present an adaptive graph-based personalized recommendation method based on co-ranking and query-based collaborative diffusion. By utilizing the unique network structure of n-partite heterogeneous graph, we attempt to address the problem of personalized recommendation in a two-layer ranking process with the help of reasonable measure of high and low order relationships and analyzing the representation of user’s preference in the graph. The experiments show that this algorithm can outperform the traditional CF methods and achieve competitive performance compared with many model-based and graph-based recommendation methods, and have better scalability and flexibility.

Rank aggregation, which combines multiple individual rank lists toobtain a better one, is a fundamental technique in various applications such as meta-search and recommendation systems. Most existing rank aggregation methods blindly combine multiple rank lists with possibly considerable noises, which often degrades their performances. In this paper, we propose a new model for robust rank aggregation (RRA) via matrix learning, which recovers a latent rank list from the possibly incomplete and noisy input rank lists. In our model, we construct a pairwise comparison matrix to encode the order information in each input rank list. Based on our observations, each comparison matrix can be naturally decomposed into a shared low-rank matrix, combined with a deviation error matrix which is the sum of a column-sparse matrix and a row-sparse one. The latent rank list can be easily extracted from the learned low-rank matrix. The optimization formulation of RRA has an element-wise multiplication operator to handle missing values, a symmetric constraint on the noise structure, and a factorization trick to restrict the maximum rank of the low-rank matrix. To solve this challenging optimization problem, we propose a novel procedure based on the Augmented Lagrangian Multiplier scheme. We conduct extensive experiments on meta-search and collaborative filtering benchmark datasets. The results show that the proposed RRA has superior performance gain over several state-of-the-art algorithms for rank aggregation.

In photo-sharing websites and in social networks, photographs are most often browsed as a sequence: users who view a photo are likely to click on those that follow. The sequences of photos (which we call photostreams), as opposed to individual images, can therefore be considered to be very important content units in their own right. In spite of their importance, those sequences have received little attention even though they are at the core of how people consume image content. In this paper, we focus on photostreams. First, we perform an analysis of a large dataset of user logs containing over 100 million pageviews, examining navigation patterns between photostreams. Based on observations from the analysis, we build a stream transition graph to analyze common stream topic transitions (e.g., users often view “train” photostreams followed by “firetruck” photostreams). We then implement two stream recommendation algorithms, based on collaborative filtering and on photo tags, and report the results of a user study involving 40 participants. Our analysis yields interesting insights into how people navigate between photostreams, while the results of the user study provide useful feedback for evaluating the performance and characteristics of the recommendation algorithms.

Current travel recommendation systems are helpful in addressing a traveler's information needs to certain extent, however, most of them fail to factor in the user in their recommendations. TripEneer proposes travel recommendations to a traveler by keeping the user preferences and constraints as first class citizens. We present an intuitive UI for helping users plan their travel trips quickly and easily. In the current demo we present various global and user-specific ranking models used for recommending travel destinations. Our preliminary evaluation showed that the users found the personalized recommendations, based on the user model, most useful.

Two main approaches to using social network information in recommendation have emerged: augmenting collaborative filtering with social data and algorithms that use only ego-centric data. We compare the two approaches using movie and music data from Facebook, and hashtag data from Twitter. We find that recommendation algorithms based only on friends perform no worse than those based on the full network, even though they require much less data and computational resources. Further, our evidence suggests that locality of preference, or the non-random distribution of item preferences in a social network, is a driving force behind the value of incorporating social network information into recommender algorithms. When locality is high, as in Twitter data, simple k-nn recommenders do better based only on friends than they do if they draw from the entire network. These results help us understand when, and why, social network information is likely to support recommendation systems, and show that systems that see ego-centric slices of a complete network (such as websites that use Facebook logins) or have computational limitations (such as mobile devices) may profitably use ego-centric recommendation algorithms.

Learning sentiment models from short texts such as tweets is a notoriously challenging problem due to very strong noise and data sparsity. This paper presents a novel, collaborative filtering-based approach for sentiment prediction in twitter conversation threads. Given a set of sentiment holders and sentiment targets, we assume we know the true sentiments for a small fraction of holder-target pairs. This information is then used to predict the sentiment of a previously unknown user towards another user or an entity using collaborative filtering algorithms. We validate our model on two Twitter datasets using different collaborative filtering techniques. Our preliminary results demonstrate that the proposed approach can be effectively used in twitter sentiment prediction, thus mitigating the data sparsity problem.

Consider a movie recommendation system where apart from the ratings information, side information such as user's age or movie's genre is also available. Unlike standard matrix completion, in this setting one should be able to predict inductively on new users/movies. In this paper, we study the problem of inductive matrix completion in the exact recovery setting. That is, we assume that the ratings matrix is generated by applying feature vectors to a low-rank matrix and the goal is to recover back the underlying matrix. Furthermore, we generalize the problem to that of low-rank matrix estimation using rank-1 measurements. We study this generic problem and provide conditions that the set of measurements should satisfy so that the alternating minimization method (which otherwise is a non-convex method with no convergence guarantees) is able to recover back the {\em exact} underlying low-rank matrix. In addition to inductive matrix completion, we show that two other low-rank estimation problems can be studied in our framework: a) general low-rank matrix sensing using rank-1 measurements, and b) multi-label regression with missing labels. For both the problems, we provide novel and interesting bounds on the number of measurements required by alternating minimization to provably converges to the {\em exact} low-rank matrix. In particular, our analysis for the general low rank matrix sensing problem significantly improves the required storage and computational cost than that required by the RIP-based matrix sensing methods \cite{RechtFP2007}. Finally, we provide empirical validation of our approach and demonstrate that alternating minimization is able to recover the true matrix for the above mentioned problems using a small number of measurements.