Towards a balanced annotated Low Saxon dataset for diachronic investigation of dialectal variation

Towards a balanced annotated Low Saxon dataset for diachronic investigation of dialectal variation Janine Siewert University of Helsinki janine.siewert@helsinki.fi Yves Scherrer University of Helsinki yves.scherrer@helsinki.fi J¨org Tiedemann University of Helsinki jorg.tiedemann@helsinki.fi Abstract We describe a new annotated dataset for Low Saxon with the intention to complement ex- isting corpora. This corpus covers the period from the 15th to the 21st century and is anno- tated with PoS and morphosyntactic tags as well as century and region information. This dataset will be used for diachronic dialecto- metry, but can lend itself to other NLP tasks as well. The target size is around 2000 sentences per dialect and century and at the time of writ- ing, 798 texts have been selected for inclusion in the corpus. They will be gradually added as the annnotation progresses. 1 Introduction We present a dataset for Low Saxon,1 a Germanic minority language spoken by roughly five mil- lion people in Northern Central Europe (Moseley, 2010). Despite its relatively large number of speak- ers, there are hardly any annotated corpora for this language, hampering corpus-based research into more modern varieties and causing a lack of well- functioning NLP tools. The dataset is part of our research into the di- achronic development of the internal variation in Low Saxon and builds upon the Reference Corpus Middle Low German/Low Rhenish (1200-1650) (ReN-Team, 2019) (henceforth ReN) and the LSDC dataset (Siewert et al., 2020) attempting to fill the gap between them. Therefore, it covers both his- torical and contemporary Low Saxon dialects from the Veluwe in the western corner of the language area to the Lower-Prussian dialects in the east. Our ultimate goal with this new dataset is to perform analyses of the internal variation within Low Saxon and its change over time. Questions 1Also called Low German, referring here to the varieties protected under the European Charter for Regional and Mi- nority Languages as Nedersaksisch in the Netherlands and Niederdeutsch in Germany as well as extinct eastern varieties. Figure 1: The Low Saxon dialects to be covered in the corpus. of interest are, for instance, the frequency and geo- graphical spread of features like two-part conjunc- tions (d˜ar¨umme dat, ˆer dat), double negation, and whether the perfect tense of modal verbs requires the main verb to occur in the infinitive or the perfect participle. These relate to the larger topic of inter- dialectal contact and how stable syntactic structures are when the speaker community is under constant exposure to a closely related more prestigious lan- guage and under pressure of language shift. 2 Background Low Saxon belongs to the western branch of the Germanic languages and is traditionally spoken mainly in Northern Germany and the North-Eastern Netherlands with official recognition in both coun- tries. The eastern dialects Pomeranian (POM) and Low Prussian (NPR) shown in Figure 1 were spo- ken in these regions prior to WW II.2 When the Hanseatic League lost its status in the 16th and 17th century, the Middle Low Saxon lit- erary language was replaced …

Content truncated to 500 words