ELITR Non-Native Speech Translation at IWSLT 2020

Proceedings of the 17th International Conference on Spoken Language Translation (IWSLT), pages 200–208 July 9-10, 2020. c⃝2020 Association for Computational Linguistics https://doi.org/10.18653/v1/P17 200 ELITR Non-Native Speech Translation at IWSLT 2020 Dominik Mach´aˇcek† and Jon´aˇs Kratochv´ıl† and Sangeet Sagar† and Mat´uˇs ˇZilinec† and Ondˇrej Bojar† and Thai-Son Nguyen‡ and Felix Schneider‡ and Philip Williams∗ and Yuekun Yao∗ †Charles University, ‡Karlsruhe Institute of Technology, ∗University of Edinburgh †{surname}@ufal.mff.cuni.cz, except jkratochvil@ufal.mff.cuni.cz, ‡{firstname.lastname}@kit.edu, ∗pwillia4@inf.ed.ac.uk,yyao2@exseed.ed.ac.uk Abstract This paper is an ELITR system submission for the non-native speech translation task at IWSLT 2020. We describe systems for of- fline ASR, real-time ASR, and our cascaded approach to offline SLT and real-time SLT. We select our primary candidates from a pool of pre-existing systems, develop a new end-to- end general ASR system, and a hybrid ASR trained on non-native speech. The provided small validation set prevents us from carrying out a complex validation, but we submit all the unselected candidates for contrastive eval- uation on the test set. 1 Introduction This paper describes the submission of the EU project ELITR (European Live Translator)1 to the non-native speech translation task at IWSLT 2020 (Ansari et al., 2020). It is a result of a collabora- tion of project partners Charles University (CUNI), Karlsruhe Institute of Technology (KIT), and Uni- versity of Edinburgh (UEDIN), relying on the in- frastructure provided to the project by PerVoice company. The non-native speech translation shared task at IWSLT 2020 complements other IWSLT tasks by new challenges. Source speech is non-native English. It is spontaneous, sometimes disfluent, and some of the recordings come from a particu- larly noisy environment. The speakers often have a significant non-native accent. In-domain train- ing data are not available. They consist only of native out-domain speech and non-spoken parallel corpora. The validation data are limited to 6 man- ually transcribed documents, from which only 4 have reference translations. The target languages are Czech and German. The task objectives are quality and simultaneity, unlike the previous tasks, which focused only on 1http://elitr.eu the quality. Despite the complexity, the resulting systems can be potentially appreciated by many users attending an event in a language they do not speak or having difficulties understanding due to unfamiliar non-native accents or unusual vocabu- lary. We build on our experience from the past IWSLT and WMT tasks, see e.g. Pham et al. (2019); Nguyen et al. (2017); Pham et al. (2017); Wetesko et al. (2019); Bawden et al. (2019); Popel et al. (2019). Each of the participating institutions has offered independent ASR and MT systems trained for various purposes and previous shared tasks. We also create some new systems for this task and de- ployment for the purposes of the ELITR project. Our short-term motivation for this work is to con- nect the existing systems into a working cascade for SLT and evaluate it empirically, end-to-end. In the long-term, we want to advance state of the art in non-native speech translation. 2 Overview of Our Submissions This paper is a joint report for two primary sub- …

Content truncated to 500 words