TGermaCorp – A (Digital) Humanities Resource for (Computational) Linguistics

TGermaCorp – A (Digital) Humanities Resource for (Computational) Linguistics Andy L¨ucking, Armin Hoenen, Alexander Mehler Goethe University Frankfurt Text Technology Lab luecking@em.uni-frankfurt.de, hoenen@em.uni-frankfurt.de, mehler@em.uni-frankfurt.de Abstract TGermaCorp is a German text corpus whose primary sources are collected from German literature texts which date from the sixteenth century to the present. The corpus is intended to represent its target language (German) in syntactic, lexical, stylistic and chronological di- versity. For this purpose, it is hand-annotated on several linguistic layers, including POS, lemma, named entities, multiword expressions, clauses, sentences and paragraphs. In order to introduce TGermaCorp in comparison to more homogeneous corpora of contemporary everyday language, quantitative assessments of syntactic and lexical diversity are provided. In this respect, TGermaCorp contributes to establishing characterising features for resource descriptions, which is needed for keeping track of a meaningful comparison of the ever-growing number of natural language resources. The assessments confirm the special role of proper names, whose propagation in text may influence lexical and syntactic diversity measures in rather trivial ways. TGermaCorp will be made available via hucompute.org. Keywords: German literature resource, language diversity, corpus characteristics, linguistic annotation 1. Introduction TGermaCorp is a digital humanities resource build around German literature texts from several centuries.1 The pri- mary texts are annotated on four levels: Firstly, the parts of speech are tagged according to the STTS (Schiller et al., 1999). Secondly, each token is assigned to its lemma. Thirdly, proper names are classified according to the kind of their referent (e.g., person or institution). Fourthly, clauses, sentences, paragraphs and headings are explicitly marked. Fifthly, multiword expressions are identified. All annota- tions have been carried out by linguistically trained annota- tors. One characteristic of TGermaCorp is the composition of its primary sources: TGermaCorp is designed in view of cap- turing the lexical and morpho-syntactic varieties of written German as exhibited in German-speaking literature. Thus, TGermaCorp complements corpora that address the homo- geneous style of mainly contemporary German (e.g., news- paper texts) like TIGER (TIGER project, 2003), (Brants et al., 2004), DeReKo (DEREKO project, 1999), (Dipper et al., 2002), or the Huge German Corpus (Stuttgart, 2010), (Schiller et al., 1999), as well as individual language re- sources like the Kant Korpus (Schmitz and Stark, 2008) (Lenders and Schmitz, 2007). Given these characteristics, TGermaCorp aims at applications and investigations within the field of Digital Humanities and therefore is located in the low-resource intersection area between computational linguistics and the study of literature. The corpus will be made publicly available via www.hucompute.org un- der the Creative Commons license CC BY-SA 3.0 DE. 2. Qualifying TGermaCorp What are the primary sources of TGermaCorp and in which way have they been collected? What is the size of TG? Are the POS annotations of TGermaCorp reliable? These questions are addressed subsequently. 1‘TGermaCorp’ is not an acronym though being related to “text technology”, “German” and “corpus”. 2.1. Composition of Primary Sources In order to obtain a selection of literary texts, we followed canonical advices as documented in three expert sources: the LiMoST database for motives …

Content truncated to 500 words