Ana Paula Appel

Large-Scale Multi-agent-Based Modeling and Simulation of Microblogging-Based Online Social Network

Online Social Networks (OSN) are self-organized systems with emergent behavior from the individual interactions. Microblogging services in OSN, like Twitter and Facebook, became extremely popular and are being used to target marketing campaigns. Key known issues on this targeting is to be able to predict human behavior like posting, forwarding or replying a message with regard to topics and sentiments, and to analyze the emergent behavior of such actions. To tackle this problem we present a method to model and simulate interactive behavior in microblogging OSN taking into account the users sentiment. We make use of a stochastic multi-agent based approach and we explore Barack Obama’s Twitter network as an egocentric network to present the experimental simulation results. We demonstrate that with this engineering method it is possible to develop social media simulators using a bottom-up approach (micro level) to evaluate the emergent behavior (macro level) and our preliminary results show how to better tune the modeler and the sampling and text classification impact on the simulation model.

Prophet -- A Link-Predictor to Learn New Rules on NELL

Link prediction is a task that in graph-based data models, as well as, in complex networks not only to predict edges that will appear in a near future but also to find missing edges. NELL is a never ending language learner system that has the ability to continuously learn to extract structured information from unstructured text (fetched from web pages) and map this information to a continuously growing knowledge base. NELLs knowledge base can be seen as a complex network, allowing us to apply graph mining techniques to extract new knowledge and enhance the system performance. In this paper we present Prophet, a link prediction component that can be connected to NELL allowing the it to infer new rules and misplaced connections among nodes, thus, helping the never-ending system to learn more and better each day. We also show that Prophet can extract new knowledge that cannot be obtained using traditional first order rule extraction procedures.

Building Socially Connected Skilled Teams to Accomplish Complex Tasks

Solving todays problems demands more than the effort of an individual, however, brilliant mind. Collaboration and team work are fundamental skills for tackling such problems. The ability of team members to work together and communicate with one another thus becomes an uppermost concern. In this context, to assemble an effective team requires an approach that goes beyond the analysis of individual skills. This paper proposes and examines the problem that takes into account different skill attributes and social ties to build an interconnected team. Our proposed solution is evaluated by means of building one team to defeat an opposite team defined in the same social network. Our experimental results show that our algorithms produces meaningful socially collaborative skilled teams.

A simulation-based approach to analyze the information diffusion in microblogging online social network

In this paper we propose a stochastic multi-agent based approach to analyze the information diffusion in Microblogging Online Social Networks (OSNs). OSNs, like Twitter and Facebook, became extremely popular and are being used to target marketing campaigns. Key known issues on this targeting is to be able to predict human behavior like posting a message with regard to some topics, and to analyze the emergent behavior of such actions. We explore Barack Obamas Twitter network as an egocentric network to present our simulation-based approach and predictive behavior modeling. Through experimental analysis, we evaluated the impact of inactivating both Obama and the most engaged users, aiming at understanding the influence of those users that are the most likely to disseminate information over the network.

Characterizing Time-Bounded Incident Management Systems

In this paper we propose an analytical tool, named Workload Profile Chart (WPC), to characterize the performance and quality of time-bounded incident management (TBIM) systems. Based on the normalization of incident assignment and resolution durations by their respective service level agreement (SLA), the method computes and plots the spreading of incidents on a log-log chart. We claim that this visual representation helps characterizing the performance of TBIM systems and diagnosing major issues such as resource and skill allocation, abnormal behavior, ticket characteristics, and the like. We also propose the WPC Inspection method which formalizes the process of using a WPC to characterize specific issues of TBIMs. The proposed method can be used to identify classes of problems for automated resolution or assignment, to determine resources and skills needed, and to reach a balance between productivity and quality. In addition to an in-depth description of the method, this paper presents its application in the characterization of four service organizations of a large IT service factory. As result, we are able to show aspects and characteristics of these service organizations that for the most part went unnoticed before. We also carried out an initial qualitative validation which provided evidence of the WPCs ability to accurately characterize TBIM systems.

Autonomously reviewing and validating the knowledge base of a never-ending learning system

The amount of information available on the Web has been increasing daily. However, how one might know what is right or wrong? Does the Web itself can be used as a source for verification of information? NELL (Never-Ending Language Learner) is a computer system that gathers knowledge from Web. Prophet is a link prediction component on NELL that has been successfully used to help populate its knowledge database. However, during link prediction task performance Prophet classify some edges as misplaced edges, that is, edges that we can not assure if they are right or not. In this paper we use the Web itself, using question answer (QA) systems, as a Prophet extension to validate these edges. This is an important issue when working with a self-supervised system where inserted errors might be propagate and generate dangerous concept drifting.

GraPhys: Understanding Health Care Insurance Data through Graph Analytics.

Healthcare insurance data represent a rich source of information and has the potential to contribute significantly in guiding business decision making. In this work we present GraPhys, a Graph Analysis platform designed for exploration, visualization and analysis of healthcare insurance data and its corresponding metadata. By taking advantage of relationships contained in healthcare claims data, we are able to apply Graph Analytics methods and algorithms in order to devise useful business metrics to guide data analysis and exploration. Our tool focuses in better understanding physicians, patients and their practices. We illustrate our approach by demonstrating two use cases where we show how graph analytics metrics, combined with other data, may lead to useful insights not directly available to traditional Business Analytics.

Reaction times for user behavior models in microblogging online social networks

Online Social Networks (OSNs) have, in recent years, emerged as a new way to communicate, diffuse information, coordinate people, establish relationships, among other possibilities. In this context, being able to understand and predict how users behave and developing appropriate models is a key problem to work with OSNs, concerning from marketing campaigns to social movements, for example. Twitter, for instance, was a heavily explored tool in Obamas 2012 election. In this paper, we explore Obamas Twitter network and model its users behavior, applying a stochastic multi-agent based simulation to reproduce the observed data. We study the effects of different time discretizations when applying a first order Markov Model to learn the user behavior and determine that, for Obamas egocentric network, users present a short reaction time to received messages.

Data-Driven Analytical Tools for Characterization of Productivity and Service Quality Issues in IT Service Factories

In this article, we propose an analytical tool, named the Workload Profile Diagnosis (WPD) method, to evaluate the performance and quality of incident management (IM) systems in information technology (IT) service factories. Based on the normalization of ticket assignment delay and resolution time by their respective service-level agreement, the method computes and plots the spreading of ticket data on a log-log chart. By comparing the actual and desired distribution values in specific areas, the WPD method diagnoses specific problems and issues in the performance of IM systems such as resource and skill allocation and abnormal behavior, and identifies opportunities for automated resolution or assignment of tickets, increases or decreases in the resources and skills needed, and ultimately aims to strike a better balance between productivity and service quality. In addition to an in-depth description of the WPD method, this article presents its application in the diagnostics of four service pools of a large IT service factory. An empirical study conducted in the IT service factory shows that most of the problems identified by the WPD method were indeed present in the service pools, therefore providing evidence of the validity of the WPD method. We conclude discussing how managers can use the method to detect and evaluate transformational opportunities to increase productivity and service quality in a systematic manner.

Building a Question-Answering Corpus Using Social Media and News Articles

Is it possible to develop a reliable QA-Corpus using social media data? What are the challenges faced when attempting such a task? In this paper, we discuss these questions and present our findings when developing a QA-Corpus on the topic of Brazilian finance. In order to populate our corpus, we relied on opinions from experts on Brazilian finance that are active on the Twitter application. From these experts, we extracted information from news websites that are used as answers in the corpus. Moreover, to effectively provide rankings of answers to questions, we employ novel word vector based similarity measures between short sentences (that accounts for both questions and Tweets). We validated our methods on a recently released dataset of similarity between short Portuguese sentences. Finally, we also discuss the effectiveness of our approach when used to rank answers to questions from real users.