The USC Institute of Information Sciences is a global leader in the research and development of advanced information processing, computer and communication technologies.
ISI, an offshoot of the USC Viterbi School of Engineering, is one of the largest and most successful university-affiliated computing institutes in the United States. The institution's work ranges from theoretical foundational research, such as core engineering and computer science discoveries, to applied research and development, such as the design and modeling of innovative prototypes anddevices

Research Areas

· Machine Learning and Applications
· Natural Language Processing
· Knowledge Graphs
· Science Data Analysis and Discovery
· Multi-modal Understanding
· Commen Sence Representation and Reasoning
· Computational Social Science
· AI Fairness



The Multi-modal Open World Grounded Learning and Inference (MOWGLI) project is our commonsense reasoning project. The goal is to build a system that can answer a wide range of common sense questions posed using either an image or natural language, about everyday intuitive phenomena such as abduction, analogy, causality, agency, physics, and social interactions.


Modern knowledge graphs (KGs) are built using a combination of structured data, crowd sourced contributions, and the output of information extraction from documents, images and video. Modern KG-reasoning methods emphasize statistical reasoning based on deep learning. The objective of KGTK is to build the "Scikit-learn" of KGs, a comprehensive library of tools and methods to make it easy to create, integrate, denoise, reason and query KGs to build interesting applications that leverage vast amounts of knowledge.


In the age of Photoshop and social media, a picture may still be worth a thousand words, but not all those words are true all the time. The goal of our DiSPARITY research effort is to create techniques that can assess the truthfulness, or integrity, of an image or a video clip—thereby allowing users to determine whether and how much to rely upon the specific visual information at hand.  


The ISI Datamart project is building technology to create the largest publicly available knowledge graph to power data-driven models in a wide variety of domains. At the core of Datamart is Wikidata, a publicly available knowledge graph that already contains over 60 million entities, and thanks to a vibrant community of contributors it is growing at a rate of about 5 edits per second. 


Information sources such as relational databases, spreadsheets, XML, JSON, and Web APIs contain a tremendous amount of structured data that can be leveraged to build and augment knowledge graphs. However, they rarely provide a semantic model to describe their contents. We present a novel approach that exploits the knowledge from a domain ontology and the semantic models of previously modeled sources to automatically learn a rich semantic model for a new source.


The goal of BOKN is to create a public resource containing knowledge about businesses, their products, and their patents as well as the relationships between them, such as customer, competitor or supplier. We envision that BOKN will be useful for entrepreneurs and innovators starting small businesses trying to understand the existing competitive landscape, allow regulators find corporate malfeasance.