German Radio Interviews: The GRAIN Release of the SFB732 Silver Standard Collection

German Radio Interviews: The GRAIN Release of the SFB732 Silver Standard Collection Katrin Schweitzer, Kerstin Eckart, Markus G¨artner, Agnieszka Falenska, Arndt Riester, Ina R¨osiger, Antje Schweitzer, Sabrina Stehwien, Jonas Kuhn University of Stuttgart, Institute for Natural Language Processing (IMS) Pfaffenwaldring 5B, 70569 Stuttgart, Germany {firstname.lastname}@ims.uni-stuttgart.de Abstract We present GRAIN (German RAdio INterviews) as part of the SFB732 Silver Standard Collection. GRAIN contains German radio interviews and is annotated on multiple linguistic layers. The data has been processed with state-of-the-art tools for text and speech and therefore represents a resource for text-based linguistic research as well as speech science. While there is a gold standard part with manual annotations, the (much larger) silver standard part (which is growing as the radio station releases more interviews) relies completely on automatic annotations. We explicitly release different versions of annotations for the same layers (e.g. morpho-syntax) with the aim to combine and compare multiple layers in order to derive confidence estimations for the annotations. Therefore, parts of the data where the output of several tools match can be considered clear-cut cases, while mismatches hint at areas of interest which are potentially challenging or where rare phenomena can be found. Keywords: corpus, silver standard, text-speech-interface 1. Introduction Over the years, many resources for natural language such as tools or corpora have been developed and are used in research and applications. Processing is usually focused on a specific type of primary data, which we call canoni- cal data for the respective tool type or branch of research. However, with a large set of efficient tools available, the time has come to make the step beyond canonical data, at the same time connecting research branches such as those based on text and speech. The SFB732 Silver Standard Col- lection is intended to serve as a resource in this respect. It is a non-static collection, i.e. more and different primary data are added over time, as are different annotation layers. This paper focuses on the first released part of the data set, containing richly annotated German radio interviews, made available for research and education in the GRAIN release of the collection. Different research groups contributed in its curation, among them were groups usually working pri- marily with written language data as well as groups who usually focus on speech, i.e. the tools applied are state-of- the art tools from both disciplines. Thereby, the resource can provide a starting point for joint exploration of speech and text. Moreover, bringing together tools from different research branches can result for some tools in encountering data that is non-canonical. For instance, parsers trained on newspaper text are to deal with incomplete sentences when being presented with spoken language data. In the remainder of the paper, we first present the idea of a Silver Standard, relying on automatic annotations, then we describe the GRAIN release: the primary data is described in Section 3, details about the existing annotations can be found in Section 4. Section 5 outlines how the workflow of creating the …

Content truncated to 500 words