A Multilingual Speech Resource: The Nordic Dialect Corpus

A multilingual speech resource: The Nordic Dialect Corpus * Janne Bondi Johannessen, Joel Priestley, Anders Nøklestad The Text Laboratory, Department of Linguistics and Nordic Studies, University of Oslo, P.O.Box 1102, Blindern, N-0317 Oslo, Norway {jannebj, joel.priestley, noklesta}@iln.uio.no Abstract. This paper describes the Nordic Dialect Corpus, a corpus that consists of transcribed spoken dialects, with sound and video, from five North European languages (Danish, Faroese, Finnish, Icelandic, Norwegian and Swedish). The paper focuses on recent developments that have been added as a result of wishes expressed by the linguist users. These include map views of various selections of search results, English translations of every dialect concordance, and search possibiities and presentation of both orthographic and phonetic transcriptions. Keywords: Nordic languages, corpus, multilingual, multimodal, speech 1 Introduction We present a corpus that is one of the major outcomes of the big collaborative dialect project Scandinavian Dialect Syntax (ScanDiaSyn) and the Nordic Centre of Excellence in Microcomparative Syntax (NORMS). In this paper we will focus specifically on features that the linguist users have asked for. The project has members from universities representing dialects from five languages in six countries in Northern Europe (Denmark, Faroe Islands, Finland, Iceland, Norway and Sweden). The majority of the project members (as well as present and future users) are linguists and dialectologists, and have little interest in and knowledge of technical matters; their main focus is dialectological research and data collection. The development of the Nordic Dialect Corpus (Johannessen et al. 2009; Johannessen et al. 2010), which we focus on in this paper, and the Nordic Syntactic Judgments Database (Lindstad et al. 2009) are the technical responsibility of the Text Laboratory, University of Oslo. When planning and constructing the corpus, we had to pay special attention to user-friendliness, but also to the requirements of the users w.r.t. search options and results handling. There are a number of factors that are challenging with respect to constructing search options and results handling in this corpus: • there are five different standard orthographies corresponding to the five standard languages (Danish, Faroese, Icelandic, Norwegian and Swedish) • the corpus contents consists of transcribed speech • some of the recordings have a double set of transcriptions – orthographic and phonetic • transcriptions should be linked to audio and video and presented nicely • the corpus should be tagged, needing five spoken language taggers, but the tagsets cannot be the same due to linguistic differences • where possible, the same tags should refer to the same types of entity in all the languages • metadata on informants should be usable as filters in search (age, sex, place) * The technical development of the corpus and the Norwegian recordings have been generously funded by NordForsk, NOS-HS, The University of Oslo and the Norwegian Research Council. In addition, the national research councils of Denmark, Iceland and Sweden have funded the national projects DanDiaSyn, IceDiaSyn, Swedia 2000 and SweDiaSyn, responsible for recordings that are now part of the corpus. Copyright 2010 by Janne Bondi Johannessen, Joel Priestley …

Content truncated to 500 words