Lecture by Dr. Sriram - Using Declarative Analytics to Unlock Value from Big Data

With (big) data seen as the fundamental natural resource of the information age, there is tremendous innovation on in systems and technologies for unearthing value from this resource. A variety of proprietary and open source technologies are being developed to acquire, integrate, and analyze noisy multi-modal data sources at scale, from within and outside the enterprise. Analytic engines for information extraction, text analytics, large-scale data integration, and machine learning are particularly critical building blocks in this process.

To explain the growing importance of Data Analytics, Dr. Sriram Raghavan, Head of Information & Analytics of IBM Research-India, delivered a talk on Using Declarative Analytics to Unlock Value from Big Data on November 14th, 2014 at IDRBT. He demonstrated how data collected from various sources like company websites, public domains like social networking sites, blogs, etc., is integrated into machine language to deduce meaningful information and make right decisions in business. With more companies banking on social media for taking a 360° view of customer behaviour (emotions, sentiments, etc.), he explained how unstructured data could be structured using analytics and highlighted the need to democratize analytics. He also:

  • Drew an analogy with the development and evolution of relational database systems and argued that some key design principles that drove the success of those systems also have value for this new generation of analytic engines.
  • Demonstrated how these engines have been deployed for a wide variety of practical applications – from social media analysis and risk modeling to IT operational analysis, telecom network analysis, and warranty analysis.

He also focused attention on a few challenges in terms of data quality, performance and productivity when using analytics for unstructured data. The lecture concluded with a demo on how information is extracted from unstructured data using AQL.