CoVoST: A Diverse Multilingual Speech-To-Text Translation Corpus

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 4197–4203 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 4197 CoVoST: A Diverse Multilingual Speech-To-Text Translation Corpus Changhan Wang, Juan Pino, Anne Wu*, Jiatao Gu Facebook AI {changhan, juancarabina, annewu, jgu}@fb.com Abstract Spoken language translation has recently witnessed a resurgence in popularity, thanks to the development of end-to-end models and the creation of new corpora, such as Augmented LibriSpeech (Kocabiyikoglu et al., 2018) and MuST-C (Di Gangi et al., 2019). Existing datasets involve language pairs with English as a source language, involve very specific domains or are low resource. We introduce CoVoST, a multilingual speech-to-text translation corpus from 11 languages into English, diversified with over 11,000 speakers and over 60 accents. We describe the dataset creation methodology and provide empirical evidence of the quality of the data. We also provide initial benchmarks, including, to our knowledge, the first end-to-end many-to-one multilingual models for spoken language translation. CoVoST is released under CC0 license and free to use. We also provide additional evaluation data derived from Tatoeba under CC licenses. Keywords: corpus, multilingual speech-to-text translation, spoken language translation, end-to-end model, CC licensed 1. Introduction End-to-end speech-to-text translation (ST) has attracted much attention recently (Berard et al., 2016; Duong et al., 2016; Weiss et al., 2017; Bansal et al., 2017; B´erard et al., 2018) given its simplicity against cascading automatic speech recognition (ASR) and machine translation (MT) systems. The lack of labeled data, however, has become a major blocker for bridging the performance gaps between end-to-end models and cascading systems. Several corpora have been developed in recent years. Post et al. (2013) introduced a 180-hour Spanish-English ST corpus by aug- menting the transcripts of the Fisher and Callhome corpora with English translations. Di Gangi et al. (2019) created the largest ST corpus to date from TED talks but the lan- guage pairs involved are out of English only. Beilharz et al. (2019) created a 110-hour German-English ST corpus from LibriVox audiobooks. Godard et al. (2018) created a Moboshi-French ST corpus as part of a rare language documentation effort. Woldeyohannis et al. (2018) pro- vided an Amharic-English ST corpus in the tourism do- main. Boito et al. (2019) created a multilingual ST corpus involving 8 languages from a multilingual speech corpus based on Bible readings (Black, 2019). Previous work ei- ther involves language pairs out of English, very specific domains, very low resource languages or a limited set of language pairs. This limits the scope of study, including the latest explorations on end-to-end multilingual ST (In- aguma et al., 2019; Gangi et al., 2019). Our work is mostly similar and concurrent to Iranzo-Sanchez et al. (2019) who created a multilingual ST corpus from the European Par- liament proceedings. The corpus we introduce has larger speech durations and more translation tokens. It is diver- sified with multiple speakers per transcript/translation. Fi- nally, we provide additional out-of-domain test sets. In this paper, we introduce CoVoST, a multilingual ST …

Content truncated to 500 words