A CLARIN Transcription Portal for Interview Data

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 3353–3359 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 3353 A CLARIN Transcription Portal for Interview Data Christoph Draxler1, Henk van den Heuvel2, Arjan van Hessen3, Silvia Calamai4, Louise Corti5, Stefania Scagliola6 1BAS/LMU, Munich, Germany ; 2CLS/CLST, Radboud University, Nijmegen, 3University of Twente, Enschede, the Netherlands, 4University of Siena, 5University of Essex, 6University of Luxembourg, draxler@phonetik.uni-muenchen.de, h.vandenheuvel@let.ru.nl, a.j.vanhessen@utwente.nl Abstract In this paper we present a first version of a transcription portal for audio files based on automatic speech recognition (ASR) in various languages. The portal is implemented in the CLARIN resources research network and intended for use by non-technical scholars. We explain the background and interdisciplinary nature of interview data, the perks and quirks of using ASR for transcribing the audio in a research context, the dos and don’ts for optimal use of the portal, and future developments foreseen. The portal is promoted in a range of workshops, but there are a number of challenges that have to be met. These challenges concern privacy issues, ASR quality, and cost, amongst others. Keywords: automatic speech recognition, interviews, digital humanities, social sciences, research infrastructure 1. Background and Aims Interview data are cross disciplinary data. Transcription is important for many research domains. Scholars and their assistants are carrying out meticulous and time staking work to convert the audio speech stream into corresponding texts. Automatic speech recognition (ASR) has reached a performance level where, under favorable acoustic conditions, a quality of transcriptions can be achieved that is a sufficient starting point for many researchers to start subsequent (domain specific) text analysis (labelling and encoding on). An additional advantage of using ASR for transcription purposes is that the output comes with time stamps of the words locating them in the original audio stream and permitting seamless subtitling of audio and video recordings. The idea of the relevance of building a web portal that would encompass a Transcription Chain converting audio to various text formats for researchers originated in the realm of oral history. The chain would allow the correction of the resulting text and aligning it again with the original audio file. Its concept was first introduced during a CLARIN workshop in Arezzo, Italy, May 2017 for oral history researchers, dialect specialists and speech technologists, and then implemented as a OH Transcription portal by developers of the Bavarian Archive for Speech Signals (BAS) in Munich (Van den Heuvel, et al., 2019). The team behind the workshop created a website with an extensive documentation of the techniques, https://oralhistory.eu/. Since then the portal has been introduced and explained in many workshops1 for an audience that became more and more diverse in terms of scientific background, thus confirming our idea that interview data indeed are an interdisciplinary research instrument for which automatic disclosure bears immense potential (Scagliola, 2019). In our most recent workshops, we welcome colleagues from the disciplines such as (oral) history, linguistics and the 1 For an overview …

Content truncated to 500 words