Linguistic and Sociolinguistic Annotation of 17th Century Dutch Letters

Linguistic and Sociolinguistic Annotation of 17th Century Dutch Letters Marijn Schraagen, Feike Dietz, Marjo van Koppen Utrecht University, The Netherlands {M.P.Schraagen, F.M.Dietz, J.M.vanKoppen}@uu.nl Abstract Developments in the Dutch language during the 17th century, part of the Early Modern period, form an active research topic in historical linguistics and literature. To enable automatic quantitative analysis, a corpus of letters by the 17th century Dutch author and politician P.C. Hooft is manually annotated with parts-of-speech, document segmentation and sociolinguistic metadata. The corpus is developed as part of the Nederlab online research portal, which is available through the CLARIN ERIC European research infrastructure. This paper discusses the design and evaluation of the annotation effort, as well as adding new annotations to an existing annotated corpus. Keywords: Early Modern Dutch, POS tagging, sociolinguistic annotation, data integration 1. Introduction In the late 16th and 17th century, the Dutch language was subject to a series of standardization and modernization de- velopments, described in historical linguistics as the transi- tion into (Early) Modern Dutch (Donaldson, 1983; Burke, 2005). These developments include changes in spelling and vocabulary, as well as syntactic changes regarding case marking and negation. The adoption of developments dif- fers not only between authors, but also within the same au- thor (van der Wal et al., 2012; Nobels and Rutten, 2014), cf. intra-speaker variation in speech corpora (Schilling-Estes, 2002; Szmrecsanyi, 2005) or intra-author variation in En- glish literature (Leech, 1969; Busse, 2002). Variation is influenced by linguistic contexts as well as sociolinguistic factors (e.g., intended audience). To aid the study of lan- guage variation in this time period, a corpus is developed containing annotations on both the linguistic and sociolin- guistic level. This corpus is intended to be used directly as linguistic research data, to study phenomena of interest both within the corpus itself as well as in comparison to external resources. Moreover, the manual annotation effort can be leveraged as training data for automatic methods in the field of historical linguistics. 2. Corpus Development The corpus includes documents from the correspondence archive of the author and politician P.C. Hooft (1581-1647), which is available online from the Digital Library of Dutch Literature (Committee DBNL, 2015). Every document in this corpus corresponds to a single letter. Hooft is known to have been interested in studying the development of the Dutch language, and advancing this development in his own writing. The full correspondence consists of over 1300 letters with a total length of 300k tokens, from which 333 letters from the period 1600-1638 have been annotated (108k tokens in total). The boundary of 1638 is selected based on a shift in the use of negation in the work of Hooft, however the letters from this time period contain a number of other interesting linguistic phenomena as well. The annotation task was performed by a pool of eight stu- dents with a background in linguistics and/or historical lit- erature. Annotations consist of lemmas, part-of-speech tags Document characteristics Category business, personal Type regular, appendix Goal express thanks, compliment, excuse, …

Content truncated to 500 words