Pronunciation Dictionaries for the Alsatian Dialects to Analyze Spelling and Phonetic Variation

Pronunciation Dictionaries for the Alsatian Dialects to Analyze Spelling and Phonetic Variation Lucie Steiblé, Delphine Bernhard LiLPa, Université de Strasbourg, France lucie.steible@unistra.fr, dbernhard@unistra.fr Abstract This article presents new pronunciation dictionaries for the under-resourced Alsatian dialects, spoken in north-eastern France. These dictionaries are compared with existing phonetic transcriptions of Alsatian, German and French in order to analyze the relationship between speech and writing. The Alsatian dialects do not have a standardized spelling system, despite a literary history that goes back to the beginning of the 19th century. As a consequence, writers often use their own spelling systems, more or less based on German and often with some specifically French characters. But none of these systems can be seen as fully canonical. In this paper, we present the findings of an analysis of the spelling systems used in four different Alsatian datasets, including three newly transcribed lexicons, and describe how they differ by taking the phonetic transcriptions into account. We also detail experiments with a grapheme-to-phoneme (G2P) system trained on manually transcribed data and show that the combination of both spelling and phonetic variation presents specific challenges. Keywords: Alsatian, transcription, spelling variation 1. Introduction The Alsatian dialects, which belong to the High German di- alects, are still spoken by approximately 500,000 speakers in Alsace, a region in north-eastern France (INSEE et al., 1999). Being non-dominant varieties, in a French-speaking country, they are mostly considered as oral languages, and have no standardized spelling system. There are however some occasions on which these dialects are written, dating back to 1816 when the first theater play in Alsatian, Der Pfingstmontag by Jean-Georges-Daniel Arnold, was pub- lished. Despite this written production, now stretching back to two centuries, almost no computational tools exist for the Alsatian dialects, which can therefore be considered as low-resourced. Given the absence of a standardized spelling, this article aims at investigating the relations between oral and writ- ten forms. This work was performed using lexicographic and lexical resources: for one resource, we used the tran- scriptions which were already provided and for the other resources we manually transcribed recordings of native speakers illustrating the lexical entries. Letter sequences and phonemes were then automatically aligned. Finally, we evaluated the challenges for a grapheme-to-phoneme (G2P) tool facing both spelling and phonetic variation. Our ultimate objective is to provide resources for the Alsatian dialects, gain a better understanding of the phonological and spelling systems and develop tools able to deal with spelling variation in text corpora. The main contributions of this article are as follows: • We present the first pronunciation dictionaries for the Alsatian dialects, obtained by the manual transcription of several datasets ; • We investigate phonetic and spelling variation in these datasets, with a focus on the specific status of conso- nants ; • We compare various spelling systems for the Alsatian dialects, with a focus on their orthographic depth ; • We train and evaluate a G2P system based on the man- ual transcriptions. 2. Related Work Spelling variation is an …

Content truncated to 500 words