Sibel Adali

Measuring behavioral trust in social networks

Trust is an important yet complex and little understood aspect of the dyadic relationship between two entities. Trust plays an important role in the formation of coalitions in social networks and in determining how high value of information flows through the network. We present algorithmically quantifiable measures of trust based on communication behavior. We propose that trust results in likely communication behaviors which are statistically different from random communications; detecting these trust-like behaviors allows us to develop a quantitative measure of who trusts whom in the network. We develop algorithms to efficiently compute such behavioral trust and validate these measures on the Twitter network.

A multi-similarity algebra

The need to automatically extract and classify the contents of multimedia data archives such as images, video, and text documents has led to significant work on similarity based retrieval of data. To date, most work in this area has focused on the creation of index structures for similarity based retrieval. There is very little work on developing formalisms for querying multimedia databases that support similarity based computations and optimizing such queries, even though it is well known that feature extraction and identification algorithms in media data are very expensive. We introduce a similarity algebra that brings together relational operators and results of multiple similarity implementations in a uniform language. The algebra can be used to specify complex queries that combine different interpretations of similarity values and multiple algorithms for computing these values. We prove equivalence and containment relationships between similarity algebra expressions and develop query rewriting methods based on these results. We then provide a generic cost model for evaluating cost of query plans in the similarity algebra and query optimization methods based on this model. We supplement the paper with experimental results that illustrate the use of the algebra and the effectiveness of query optimization methods using the Integrated Search Engine (I.SEE) as the testbed.

Predicting Personality with Social Behavior

In this paper, we examine to which degree behavioral measures can be used to predict personality. Personality is one factor that dictates peoples propensity to trust and their relationships with others. In previous work, we have shown that personality can be predicted relatively accurately by analyzing social media profiles. We demonstrated this using public data from facebook profiles and text from Twitter streams. As social situations are crucial in the formation of ones personality, ones social behavior could be a strong indicator of her personality. Given most users of social media sites typically have a large number of friends and followers, considering only these aspects may not provide an accurate picture of personality. To overcome this problem, we develop a set of measures based on ones behavior towards her friends and followers. We introduce a number of measures that are based on the intensity and number of social interactions one has with friends along a number of dimensions such as reciprocity and priority. We analyze these features along with a set of features based on the textual analysis of the messages sent by the users. We show that behavioral features are very useful in determining personality and perform as well as textual features.

An algebra for creating and querying multimedia presentations

Abstract. Over the last few years, there has been a tremendous increase in the number of interactive multimedia presentations prepared by different individuals and organizations. In this paper, we present an algebra for creating and querying interactive multimedia presentation databases. This algebra operates on trees whose branches reflect different possible playouts of a set of presentations. The algebra not only extends all the classical relational operators to such databases, but also introduces a variety of novel operators for combining multiple presentations. As our algebra supports merging parts or all of existing presentations, this algebra can also be used as an authoring tool for creating multimedia presentations. We prove a host of equivalence results for queries in this algebra, which may be used to build query optimizers for interactive presentation databases.

A Survey on Trust Modeling

The concept of trust and/or trust management has received considerable attention in engineering research communities as trust is perceived as the basis for decision making in many contexts and the motivation for maintaining long-term relationships based on cooperation and collaboration. Even if substantial research effort has been dedicated to addressing trust-based mechanisms or trust metrics (or computation) in diverse contexts, prior work has not clearly solved the issue of how to model and quantify trust with sufficient detail and context-based adequateness. The issue of trust quantification has become more complicated as we have the need to derive trust from complex, composite networks that may involve four distinct layers of communication protocols, information exchange, social interactions, and cognitive motivations. In addition, the diverse application domains require different aspects of trust for decision making such as emotional, logical, and relational trust. This survey aims to outline the foundations of trust models for applications in these contexts in terms of the concept of trust, trust assessment, trust constructs, trust scales, trust properties, trust formulation, and applications of trust. We discuss how different components of trust can be mapped to different layers of a complex, composite network; applicability of trust metrics and models; research challenges; and future work directions.

A multimedia presentation algebra

Over the last few years, there has been a tremendous increase in the number of interactive multimedia presentations prepared by different individuals and organizations. In this paper, we present an algebra for querying multimedia presentation databases. In contrast to the relational algebra, an algebra for interactive multimedia presentations must operate on trees whose branches reflect different possible playouts of a family of presentations. The query language supports selection type operations for locating objects and presentation paths that are of interest to the user, join type operations for combining presentations from multiple databases into a single presentation, and finally set theoretic operations for comparing different databases. The algebra operations can be used to locate presentations with specific properties and also for creating new presentations by borrowing different components from existing ones. We prove a host of equivalence results for queries in this algebra which may be used to build query optimizers for interactive presentation databases.

Predicting personality with social behavior: a comparative study

In this paper, we study the problem of predicting personality with features based on social behavior. While network position and text analysis are often used in personality prediction, the use of social behavior is fairly new. Often studies of social behavior either concentrate on a single behavior or trait, or simply use behavior to predict ties that are then used in analysis of network position. To study this problem, we introduce novel features based on a person’s social actions in general, towards specific individuals in particular. We also compute the variation of these actions among all the social contacts of a person as well as the actions of friends. We show that social behavior alone, without the help of any textual or network position information, provides a good basis for personality prediction. We then provide a unique comparative study that finds the most significant features based on social behavior in predicting personality for three different communication mediums: Twitter, SMS and phone calls. These mediums offer us with social behavior from public and private contexts, containing messaging and voice call type exchanges. We find behaviors that are distinctive and normative among the ones we study. We also illustrate how behavioral features relate to different personality traits. We also show the various similarities and differences between different mediums in terms of social behavior. Note that all behavioral features are based on statistical properties of the number and the time of social actions and do not consider the textual content. As a result, they can be applied in many different settings. Furthermore, our findings show us how behavioral features can be customized to a specific medium and personality trait.

The Impact of Ranker Quality on Rank Aggregation Algorithms: Information vs. Robustness

The rank aggregation problem has been studied extensively in recent years with a focus on how to combine several different rankers to obtain a consensus aggregate ranker. We study the rank aggregation problem from a different perspective: how the individual input rankers impact the performance of the aggregate ranker. We develop a general statistical framework based on a model of how the individual rankers depend on the ground truth ranker. Within this framework, one can study the performance of different aggregation methods. The individual rankers, which are the inputs to the rank aggregation algorithm, are statistical perturbations of the ground truth ranker. With rigorous experimental evaluation, we study how noise level and the misinformation of the rankers affect the performance of the aggregate ranker. We introduce and study a novel Kendalltau rank aggregator and a simple aggregator called PrOpt, which we compare to some other well known rank aggregation algorithms such as average, median and Markov chain aggregators. Our results show that the relative performance of aggregators varies considerably depending on how the input rankers relate to the ground truth.

Extended structural balance theory for modeling trust in social networks

Modeling trust in very large social networks is a hard problem due to the highly noisy nature of these networks that span trust relationships from many different contexts, based on judgments of reliability, dependability and competence and the relationships vary in their level of strength. In this paper, we introduce a new extended balance theory as a foundational theory of trust in networks. Our theory preserves the distinctions between trust and distrust as suggested in the literature, but also incorporates the notion of relationship strength which can be expressed as either discrete categorical values, as pairwise comparisons or as metric distances. Our model is novel, has sound social and psychological basis, and captures the classical balance theory as a special case. We then propose a convergence model, describing how an imbalanced network evolves towards new balance and formulate the convergence problem of a social network as a Metric Multidimensional Scaling (MDS) optimization problem. Finally, we show how the convergence model can be used to predict edge signs in social networks, and justify our theory through experiments on real datasets.

A flexible architecture for query integration and mapping

The aim of information integration is to build sophisticated information systems by making use of the available information sources to the fullest extent and by pushing costly operations to the sources as much as possible. This is especially true when translating queries across multiple multimedia information sources that support advanced and/or similarity based queries. We propose a flexible architecture that allows users to specify a wide range of structured queries using generic and simple query constructs and connectives through a uniform query interface. These queries are translated into resource specific queries by processing rules that specify the properties of different query interfaces as well as the user preferences for evaluating queries. We show how different and multiple notions of query relaxation can be captured in this framework naturally.