Speaker Details

Gregor Sieber

Head of Consulting and Innovation @ EBCONT

Gregor started his career in information technology by completing a degree in Computational Linguistics at the University of Tübingen and in Computer Science Management at the Technical University of Vienna. Following university, Gregor spent 6 years at the Language Technologies group of the Austrian Research Institute for Artificial Intelligence (OFAI), working on a variety of international and national research projects. With a decision to switch sides from academia to industry around 7 years ago, Gregor joined EBCONT, an Austrian IT solution provider, where he manages a portfolio of projects in areas such as search, natural language processing, machine learning, big data, and other innovation areas. Gregor likes to ride his bike to work and leave the car at home; in his free time he enjoys human powered adventures that usually involve water or mountains.

Does Your Algorithm Understand Emails? Digging Into Unstructured Data with NLP and Scalable Architectures

When we look at the data produced and stored today - while exact estimates vary - there is a general consensus that large amounts of it are unstructured. For data professionals, this means that methods from Natural Language Processing or Computational Linguistics increasingly need to become part of our portfolio to tackle the challenge of finding business critical information within such data - usually without the comfort of a properly annotated training set. This talk will present some of our experience in building real-world business applications dealing with natural language documents for domains such as court cases and medical literature. It will showcase the scalable architectures and approaches we use to bootstrap machine learning models or search algorithms with minimum amounts of human annotation for tasks such as finding and linking entites and other data to create structured information, or extract knowledge from texts to build an ontology that includes causal relationships between terms.