Open Access

In the NLP literature, adapting a parser to new text with properties different from the training data is commonly referred to as domain adaptation. In practice, however, the differences between texts from different sources often reflect a mixture of domain and genre properties, and it is by no means clear what impact each of those has on statistical parsing. In this paper, we investigate how differences between articles in a newspaper corpus relate to the concepts of genre and domain and how they influence parsing performance of a transition-based dependency parser. We do this by applying various similarity measures for data point selection and testing their adequacy for creating genre-aware parsing models.

The present paper explores the change in distribution and potential function as well as the interplay of two phenomena that occur at the internal boundaries of nominal compounds, namely linking elements and hyphenation. About 40% of present-day German compounds contain a linking element, most prominently -s- (e.g. Geburt-s-ort ‘birth place’). Numerous theories have been brought forward to explain its function, two of which are examined here: It will be shown that the linking-s tends to mark morphologically complex constituents while the assumption that it prefers marked phonological words cannot be corroborated. Linked compounds in present-day German use hyphenation, a strategy that is mostly employed with graphematically or phonologically marked constituents, at a much smaller rate than unlinked compounds. In Early New High German (ENHG, 1350-1650), when the linked type arose by reanalyzing prenominal genitive attributes as first constituents of compounds, the reverse held true: Linked compounds underwent a gradual graphematic integration from separate writing into directly connected words which was partly reversed by a century of hyphenation (1650-1750). While hyphenation also occurred with unlinked compounds, the linked compounds show a striking preference with hyphenation rates reaching a peak at around 90%. It will be argued that ENHG hyphenation had the same function it has today, namely structuring constituents that are perceived as marked: The change in spelling between ENHG and today reflects the integration of a formerly syntactic and thereby marked pattern into word-formation.