Building a Time-Aligned Cross-Linguistic Reference Corpus from Language Documentation Data (DoReCo)

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 2657–2666 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 2657 Building a Time-Aligned Cross-Linguistic Reference Corpus from Language Documentation Data (DoReCo) Ludger Paschen†, François Delafontaine‡, Christoph Draxler*, Susanne Fuchs†, Matthew Stave‡, Frank Seifart† Leibniz-Zentrum Allgemeine Sprachwissenschaft†, Laboratoire Dynamique Du Langage‡, Bavarian Archive for Speech Signals* Schützenstraße 18 10117 Berlin, 14 Avenue Berthelot 69007 Lyon, Schellingstraße 3 80799 München {paschen, fuchs, seifart}@leibniz-zas.de, {matthew.stave, francois.delafontaine}@cnrs.fr, draxler@phonetik.uni-muenchen.de Abstract Natural speech data on many languages have been collected by language documentation projects aiming to preserve linguistic and cultural traditions in audiovisual records. These data hold great potential for large-scale cross-linguistic research into phonetics and language processing. Major obstacles to utilizing such data for typological studies include the non-homogenous nature of file formats and annotation conventions found both across and within archived collections. Moreover, time-aligned audio tran- scriptions are typically only available at the level of broad (multi-word) phrases but not at the word and segment levels. We report on solutions developed for these issues within the DoReCo (DOcumentation REference COrpus) project. DoReCo aims at providing time-aligned transcriptions for at least 50 collections of under-resourced languages. This paper gives a preliminary overview of the current state of the project and details our workflow, in particular standardization of formats and conventions, the addition of segmental alignments with WebMAUS, and DoReCo’s applicability for subsequent research programs. By making the data accessible to the scientific community, DoReCo is designed to bridge the gap between language documentation and linguistic inquiry. Keywords: corpus creation, endangered languages, phonetic databases 1. Introduction Temporal patterns of speech are of central interest in the cog- nitive sciences as they provide key evidence for the architec- ture underlying the human language production system in terms of its cognitive-neural and physiological-articulatory bases (Jaeger and Buz, 2017). But most current approaches draw on phonetic measurements from only less than one percent of the world’s languages, most of these (Indo-) Eu- ropean (Anand et al., 2015; Norcliffe et al., 2015). It has repeatedly been pointed out how problematic generaliza- tions based on data from such a thin slice of the human population are (Henrich et al., 2010). Specifically, previ- ous research has had few means to assess whether observed temporal patterns in speech are characteristic of human cog- nition and articulation in general or specific to just the few languages studied so far, predominantly English. For in- stance, recent claims on the neural basis for a preferred syllable rate (Assaneo and Poeppel, 2018) and on maxi- mum achievable rates (Ghitza, 2014) still await systematic testing on a broad cross-linguistic sample. DoReCo ad- dresses this problem by compiling a dataset for comparative studies of spoken language in a diverse sample of at least 50 languages.1 We identify and extract suitable data from language-documentation collections (Himmelmann, 1998) archived at established repositories, and process these into a multilingual reference corpus consisting of one million words of transcribed, translated, and time-aligned corpus data with …

Content truncated to 500 words