David Lenz

Machine Learning and Web Text Analytics for Economic Research

Abstract: Machine learning and artificial intelligence are key technologies for the next century. In this thesis I present my work using machine learning and especially naturallanguage processing to solve economic problems, with a focus on innovation economics. The contribution I am making with my co-authors is primarily of a methodological nature. I focus on data-driven solutions incorporating (web-based) textual data. Besides the incorporation of novel methods and data sources, another main goal of my research is to tighten the research-to-production gap, sometimes also called science-to-practice gap, i.e. the time it takes industry to exploit scientific research results. Another important, however more minor, theme is the usage of mass web data. Thereby, my co-authors and me contributed to the literature by providing several novel web based indicators through mass web text analysis. The de-veloped web data based indicators provide valuable advantages over conventional instruments, i.e. timeliness, granularity, frequency and low-resource usage.