I am Research Lead at the Center on Knowledge Graphs within USC's Information Sciences Institute (ISI), and Research Assistant Professor at the USC Viterbi School of Engineering. I obtained my MSc degree in Artificial Intelligence and a PhD in Natural Language Processing at the Vrije Universiteit (VU) in Amsterdam.
Research interest: I am developing robust and explainable neuro-symbolic technology with positive real-world impact, based on neural methods and high-quality knowledge. During my PhD, I worked on establishing identity of long-tail entities in text, which is challenging for AI because of their ambiguity, data sparsity, and the need for diverse implicit knowledge and its contextualization. After joining USC/ISI in late 2019, I have been focusing on the notorious challenge of commonsense reasoning in tasks like question answering, procedural reasoning, and story understanding. Our methods rely on neuro-symbolic reasoning, typically combining neural (language) models, machine learning, and knowledge graphs. Recently, I have been focusing on tailoring the developed methods to application domains, primarily intelligent traffic monitoring and healthcare.
Collaborators: I believe firmly in team work. In my PhD years, I have closely collaborated with my mentors (Piek Vossen, Marieke van Erp, Stefan Schlobach, and Frank van Harmelen), with Ed Hovy who I visited at CMU in 2017, with the LOD Laundromat/Triply team (Wouter Beek and Laurens Rietveld), and with Marten Postma. Through successful collaborations, I co-authored publications with other European researchers, including Giuseppe Rizzo, Raphael Troncy, and Ruben Izquierdo Bevia. My mentor at ISI has been Pedro Szekely, with whom I have been working closely on a wide range of topics covering knowledge graphs, language modeling, and neuro-symbolic reasoning. Together with Alessandro Oltramari and Jon Francis from Bosch Research and Kaixin Ma from CMU we have been developing methods for zero-shot Question Answering, generalizable model adaptation with lower data dependency, and procedural understanding (under review). In collaboration with Alessandro, Kaixin, Pedro, and Deborah McGuinness (RPI), we identified 13 dimensions of common sense based on existing sources. I have been collaborating with Jay Pujara on tasks involving numeric reasoning, story generation and explanation, and analogical reasoning. With Fred Morstatter we have been exploring biases in both factual and commonsense knowledge graphs. In collaboration with Xiang Ren and Peifeng Wang, we have been investigating novel neuro-symbolic methods for commonsense reasoning, based on path generation and scene imagination. Other joint projects with Pedro Szekely are centered around the Wikidata knowledge graph, focusing on its quality, link prediction, and similarity. The main backbone for these is the Knowledge Graph ToolKit (KGTK) - a comprehensive set of operations for working with modern, hyperrelational graphs like Wikidata.
For students: I enjoy working with students. In fact, many of the research results highlighted above have been guided by PhD, MSc, or undergraduate students. I have various interesting projects students can work on, including tasks like story understanding, fact verification, and question answering. I am actively hiring strong and passionate Master or PhD students for either a paid position or directed research. If your CV shows the right skills, I will send you a programming exercise and interview you to discuss projects. I meet weekly with my students to ensure steady progress and address challenges.