PROJECTS
| AI Fairness Methodology (2023-): Co-advising a master thesis that examines the trade-offs among metrics and methods aimed at mitigating AI bias. |
| Ethical NLP Technologies to Support Indigenous Communications (2022-2023): Investigated the robustness of current AI models to Brazilian indigenous languages at IBM Research Brazil. Focused on analyzing the representation of Brazilian Indigenous languages in Wikipedia, leading to publications in ICWSM'24. |
| Connecting Under Represented Minorities and Qualified Job Positions (2021): Explored the distribution of job postings among underrepresented groups, particularly Brazilian and U.S. Black Tech communities. Presented initial findings at the RDAI'21 workshop at AAAI, with further results published in WebSci'22. At IBM Research Brazil, we focus on understanding the first phase of hiring process, how job posts are disseminated among underrepresented groups, in particular Brazilian black communities. Our initial proposal was presented in the RDAI'21 workshop at AAAI. My focus was on characterizing the Brazilian and US Black Tech communities as well as their perceptions about the big tech companies. The results were published at (WebSci'22). |
| Vaccination Debate on Social Media (2021): Collaborated with UFMG to monitor Twitter discussions on COVID-19 vaccines since December 2020. Published results in Web Science'21. |
| Misinformation Dissemination (2018-present): Also, in collaboration with UFMG, investigated misinformation dissemination on WhatsApp during Brazilian Presidential elections. Published findings in WebMedia'18, WWW'19, WebScience'19, and WWW'20. |
| Hate speech on Social Networks (2018-2019): Explored the presence of hate speech content on Twitter during the 2014 and 2018 Soccer World Cup editions. Published results in WebScience 2019. |
| Conversational Agents Evaluation (2016-2017) : Research on methods to evaluate conversational agent systems, such as chatbots, focusing particularly on assessing these systems from a user perspective by observing the resultant interaction of all chatbot modules. A specialized chatbot tester tool was developed to simulate user connections to chatbots and collect interaction measures. This project has resulted in a patented tool and publications presented at Conversational UX Design CHI 2017 Workshop, IHC na prática (Brazilian Conference on HCI)*, CHI'19**, and CUI'19.
*We won best paper of the workshop. ** We won an honorable mention for this work. |
| Characterization and Popularity Prediction of Micro-Reviews (2011-2015): In the Ph.D. dissertation, an examination was conducted on user engagement with micro-reviews, particularly focusing on Foursquare tips, which present unique challenges due to their informal nature. Leveraging data from over 13 million Foursquare users, an analysis was carried out on the evolution of tip popularity over time, leading to the development of predictive models combining influencing factors. The study's key findings, including behavioral patterns impacting tip popularity, were published in WSDM'12, COSN'14, and SAC'14, with further investigation on tip popularity dynamics presented at COSN'14, and the overall research culminated in a publication in the Information Sciences journal in 2015. |
| Polarity Detection in Micro-Reviews or Tips (2012-2013): The study assessed the effectiveness of polarity classification strategies on subsets of the Foursquare dataset, employing supervised machine learning techniques and an unsupervised lexicon-based approach. Findings indicate that effective polarity classification can be achieved, even with the simpler lexicon-based method, as outlined in the publication presented at the SocInfo'13 conference. |
| Privacy Inference in Location-based Social Networks (2012-2013): Analyzed information leakage from publicly available features on Foursquare. Published results in LBSN'12 and PinSoDa'12 workshop. |
| Capacity Planning Models for Business Intelligence Workloads (2009-2010): In partnership with HP Labs, Palo Alto, our team tackled the development of analytical models for capacity planning and performance analysis in Business Intelligence workloads. These workloads, driven by queries processing large datasets, demanded intricate parallel processing solutions. Our efforts focused on optimizing intra-query pipeline parallelism to improve system efficiency and scalability. |
| Location Estimation of Mobile Sensors using Contact Traces (2006-2008): Proposed a model for estimating sensor locations in wireless networks using contact traces for master's thesis (M.Sc.). |
| Analysis of Streaming Media Distribution in Peer-to-Peer Architectures (2001-2003): Conducted experimental analysis comparing peer-to-peer and client/server approaches for distributing live streaming media. Published results in LA-WEB 2003 and as a Master thesis. |