KRAUTS: A German Temporally Annotated News Corpus

KRAUTS: A German Temporally Annotated News Corpus Jannik Str¨otgenα, Anne-Lyse Minardβ δ, Lukas Langeα, Manuela Speranzaβ, Bernardo Magniniβ α Max Planck Institute for Informatics, Saarland Informatics Campus, Saarbr¨ucken, Germany β Fondazione Bruno Kessler, Trento, Italy δ Univ Rennes, Inria, CNRS, IRISA, Rennes, France {jannik.stroetgen,llange}@mpi-inf.mpg.de, {minard,manspera,magnini}@fbk.eu Abstract In recent years, temporal tagging, i.e., the extraction and normalization of temporal expressions, has become a vibrant research area. Several tools have been made available, and new strategies have been developed. Due to domain-specific challenges, evaluations of new methods should be performed on diverse text types. Despite significant efforts towards multilinguality in the context of temporal tagging, for all languages except English, annotated corpora exist only for a single domain. In the case of German, for example, only a narrative- style corpus has been manually annotated so far, thus no evaluations of German temporal tagging performance on news articles can be made. In this paper, we present KRAUTS, a new German temporally annotated corpus containing two subsets of news documents: articles from the daily newspaper DOLOMITEN and from the weekly newspaper DIE ZEIT. Overall, the corpus contains 192 documents with 1,140 annotated temporal expressions, and has been made publicly available to further boost research in temporal tagging. Keywords: temporal tagging, corpus annotation, TIMEX3 1. Introduction Temporal tagging – the extraction and normalization of temporal expressions from texts – is an important task towards improved natural language understanding. Once temporal expressions have been detected in a text, their se- mantics can be assigned to them in a standard format so that applications can exploit not only the surface forms of temporal expressions, but also their meaning. For in- stance, applications in event / timeline extraction (Minard et al., 2015; Cornegruta and Vlachos, 2016; Spitz and Gertz, 2016), question answering (Llorens et al., 2015) and (tem- poral) information retrieval (Kanhabua et al., 2015) can ex- ploit temporal tagging output. Thus, temporal tagging has become a vibrant research area, and several new tempo- ral taggers have been made available and new strategies have been developed. However, as was shown in pre- vious work (Mazur and Dale, 2010; Str¨otgen and Gertz, 2013; Bethard et al., 2016; Tabassum et al., 2016), differ- ent types of documents pose different challenges for tempo- ral tagging such that domain-sensitive normalization strate- gies are required (Str¨otgen and Gertz, 2016). To judge the performance of temporal taggers and new methods, evalu- ations need to be performed on diverse text types, e.g., on news articles and narrative-style Wikipedia documents. In contrast to many natural language processing tasks, there has also been some effort towards multilinguality in the context of temporal tagging, e.g., research competitions were organized not only for English but covered further lan- guages such as Spanish and Italian (Verhagen et al., 2010; Caselli et al., 2014). Despite its importance, German has not been part of any of these challenges so far. In addition, HeidelTime is the only publicly available temporal tagger for German, and only narrative style corpora have been manually annotated so …

Content truncated to 500 words