A Very Low Resource Language Speech Corpus for Computational Language Documentation Experiments

A Very Low Resource Language Speech Corpus for Computational Language Documentation Experiments P. Godard∗, G. Adda∗, M. Adda-Decker†, J. Benjumea‡, L. Besacier ⋆, J. Cooper-Leavitt∗, G-N. Kouarata†, L. Lamel∗, H. Maynard∗, M. M¨uller⋄, A. Rialland†, S. St¨uker⋄, F. Yvon∗, M. Zanon-Boito ⋆ ∗LIMSI, CNRS, Universit´e Paris-Saclay, Orsay, France †LPP, CNRS, Paris, France ‡ENS, EHESS, PSL Research University, CNRS, INRIA, Paris, France ⋆LIG, UGA, G-INP, CNRS, INRIA, Grenoble, France ⋄KIT, Karlsruhe, Germany contact: pierre.godard@limsi.fr, laurent.besacier@univ-grenoble-alpes.fr Abstract Most speech and language technologies are trained with massive amounts of speech and text information. However, most of the world languages do not have such resources and some even lack a stable orthography. Building systems under these almost zero resource conditions is not only promising for speech technology but also for computational language documentation. The goal of computational language documentation is to help field linguists to (semi-)automatically analyze and annotate audio recordings of endangered, unwritten languages. Example tasks are automatic phoneme discovery or lexicon discovery from the speech signal. This paper presents a speech corpus collected during a realistic language documentation process. It is made up of 5k speech utterances in Mboshi (Bantu C25) aligned to French text translations. Speech transcriptions are also made available: they correspond to a non-standard graphemic form close to the language phonology. We detail how the data was collected, cleaned and processed and we illustrate its use through a zero-resource task: spoken term discovery. The dataset is made available to the community for reproducible computational language documentation experiments and their evaluation. Keywords: language documentation, field linguistics, spoken term discovery, word segmentation, zero resource technologies, unwritten languages. 1. Introduction Many languages will face extinction in the coming decades. Half of the 7,000 languages spoken worldwide are expected to disappear by the end of this century (Austin and Salla- bank, 2011), and there are too few field linguists to docu- ment all of these endangered languages. Innovative speech data collection methodologies (Bird et al., 2014; Blachon et al., 2016) as well as computational assistance (Adda et al., 2016; St¨uker et al., 2016) were recently proposed to help them in their documentation and description work. As more and more researches are related to computational language documentation (Duong et al., 2016; Franke et al., 2016a; Godard et al., 2016; Anastasopoulos and Chiang, 2017), there is a need of realistic corpora to fuel repro- ducible and replicable language studies at the phonetic, lex- ical and syntactic levels. To our knowledge, very few cor- pora are available for computational analysis of endangered languages.1 Our work follows this objective and presents a speech dataset collected following a real language documenta- tion scenario. It is multilingual (Mboshi speech aligned to French text) and contains linguists’ transcriptions in Mboshi (in the form of a non-standard graphemic form close to the language phonology). The corpus is also en- riched with automatic forced-alignment between speech and transcriptions. The dataset is made available to the re- search community2. This dataset is part of a larger data 1We are only aware of a …

Content truncated to 500 words