In this seminar, we look into a variety of approaches to data-driven parsing that extend standard constituency parsers in different ways:
1. We will discuss approaches that are based on grammar formalisms that are able to generate structures that are beyond context-free grammars. Examples are LCFRS and TAG. Topics in this context are grammar extraction, supertagging and various parsing methods.
2. Related to this, we will discuss approaches that are able to describe discontinuous constituents. These can be grammar-based (LCFRS) or transition based parsers with appropriate transitions such as gap or swap.
3. The course will also cover aspects of incremental, cognitively plausible approaches to constituency parsing.