German selbst-Compounds: Towards a NooJ Grammar

In this paper, we will describe the first steps taken to build a NooJ grammar able to annotate German selbst-compounds, a rather productive compound pattern that has attracted the attention of scholars in the domains of grammatical description, word formation and typology. After defining selbst‑compounds, we will review the existing NooJ German module. We will then detail the procedure to get to an in-depth description of the phenomenon through the analysis of a sample of a thousand occurrences of selbst-compounds in a huge Web-based corpus of German. We will show the first version of our grammar based on the insights obtained from analyzing the corpus data, describe its output, and evaluate its performance. Finally, we will outline the future steps needed for improving the performance of the grammar.