Using a Language Technology Infrastructure for German in order to Anonymize German Sign Language Corpus Data

Using a Language Technology Infrastructure for German in order to Anonymize German Sign Language Corpus Data Julian Bleicken, Thomas Hanke, Uta Salden, Sven Wagner Institute of German Sign Language and Communication of the Deaf, University of Hamburg Binderstraße 34, 20146 Hamburg, Germany E-mail: {julian.bleicken,thomas.hanke,uta.salden,sven.wagner}@sign-lang.uni-hamburg.de Abstract For publishing sign language corpus data on the web, anonymization is crucial even if it is impossible to hide the visual appearance of the signers: In a small community, even vague references to third persons may be enough to identify those persons. In the case of the DGS Korpus (German Sign Language corpus) project, we want to publish data as a contribution to the cultural heritage of the sign language community while annotation of the data is still ongoing. This poses the question how well anonymization can be achieved given that no full linguistic analysis of the data is available. Basically, we combine analysis of all data that we have, including named entity recognition on translations into German. For this, we use the WebLicht language technology infrastructure. We report on the reliability of these methods in this special context and also illustrate how the anonymization of the video data is technically achieved in order to minimally disturb the viewer. Keywords: sign language, corpus, named entity recognition, anonymization 1. Introduction In its first phase, the DGS Korpus project collected a corpus of conversations in DGS (German Sign Language) from 165 pairs of informants totaling in multi-camera video recordings of 825 hours (cf. Nishio et al., 2008). In the second phase, basic annotation as well as translations were added. As it is the aim of the project to provide language data for linguistic research and at the same time to contribute to the cultural heritage of the sign language community, parts of the corpus shall be made available to the public via a website. This of course raises the question as to what part of the data is to be anonymized before publication and how exactly this is to be accomplished. Sign language users not only have their hands as articulators, but other parts of the body are information channels as well, such as eyebrows, eye gaze, mouth, head movement etc. Hiding the face in order to make identification of the signer impossible therefore is not an option. The state of the art in avatar technology does not allow to faithfully reproduce video-recorded sign language data with economical limits to work invested. So there is no alternative to making the signers in the DGS corpus data fully visible and recognizable to the viewer. Fortunately, most participants are very proud of their involvement in the project and have agreed to their videos being published on the web. As our informants typically spent around seven hours on-site, being occupied with a variety of diverting tasks, many were fully engaged in the conversations and completely forgot about the surroundings, i.e. that there were filmed. This resulted in close-to-natural conversation, often revealing details about themselves or other persons not really suitable …

Content truncated to 500 words