Mind the semantic gap: How "talking semantics" can help you perform better data science

Description

Panos Alexopoulos frequently hears, “Now you’re talking semantics” from some data practitioners who believe that statistical reasoning and machine learning are all that’s needed to tackle semantic-related data science tasks, and that spending too much time in understanding and modeling those semantics is a waste of effort and resources. He argues in favor of more semantic rigorousness in data modeling and science by describing real use cases where such rigorousness helped significantly improve the effectiveness of the data-intensive applications.

You’ll explore how dissecting and measuring the different types of ambiguity improved the effectiveness of an entity recognition and disambiguation system, how specializing and contextualizing the notion of semantic relatedness helped mine and deliver more accurate domain knowledge to the semantic search system, and how digging deeper and challenging the semantics of publicly available data sources helped avoid ingesting data into the systems that would be nonuseful or even harmful. You’ll leave with a good idea of how data semantics are capable of both benefiting and harming a data science task. More importantly, you’ll learn basic semantic analysis techniques that will help you improve the effectiveness of your own projects, not by increasing data volume or algorithm complexity, but just by paying a bit more attention to the data and task semantics.