Read our feature in the December 2012 Sigmod Record.
The Data Analytics group at QCRI has built expertise focused on three core data management challenges that will enable the effective use of this growing asset class: extraction from its natural digital habitat, integration from a large and evolving number of sources, and robust cleaning processes to assure data quality and validation.
The Data Trio: Extraction, Integration, and Cleaning: Institutions and industries at a national level deal with large scale, heterogeneous data collected from large number of sources. The main challenge is a judicious use of the information within and across organizations to make informed decisions and to run operations effectively.
At QCRI, we are focusing on the interaction among three core data management challenges that will enable effective use of the continuously growing data: Information Extraction, Data and Schema Integration, and Data Cleaning.
Going beyond traditional ETL approaches, we are investigating multiple new directions, including: handling unstructured data; interleaving extraction, integration, and cleansing tasks in a more dynamic and interactive process that responds to evolving data sets and real-time decision-making constraints; and leveraging the power of human cycles to solve hard problems such as data cleaning and information integration.
Scalable Knowledge Models: Grand challenges mean big data. ‘Knowledge base’ is the term commonly used to refer to data, along with the rules and the logic that describe the information within this data. Large-scale knowledge management is a core-computing challenge due to the expensive process involved in reasoning about the data and inferring the facts and the various semantics embedded within. We focus on developing efficient knowledge representation models and semantic-aware query languages and processing engines that bring semantics to real applications. Main applications domains include media and health, where current approaches are either too expensive or fall short in delivering user needs.