CBFC: a parallel L2 speech corpus for Korean and French learners

CBFC: a parallel L2 speech corpus for Korean and French learners Hiyon Yoo*, Inyoung Kim+ *Laboratoire de Linguistique formelle CNRS 7110, Université Paris Diderot, Place Paul Ricoeur, Paris 75013, France + Naver Labs Europe, 6 chemin de Maupertuis, Meylan 38240, France *yoo@linguist.univ-paris.diderot.fr , +inyoung.kim@naverlabs.com Abstract In this paper, we present the design of a bilingual corpus of French learners of Korean and Korean learners of French using the same experimental design. This language resource contains mainly speech data, gathered among learners with different proficiency levels and in different speaking contexts (read and spontaneous speech). We aim at providing a translated and annotated corpus to the scientific community which can be used for a large array of purposes in the field of theoretical but also applied linguistics. Keywords: L2 corpora, Korean, French, phonetics, prosody, phonetic segmentation, alignment 1. Introduction In the last years, especially with the development of new methodologies and new technologies in corpus and computational linguistics, language corpora have become more and more common and needed in linguistic research. Corpora vary a lot in size, uses or presentations but it appears clearly that doing research using linguistic corpora brings along new questions but also new analyses of linguistic phenomena. Recently, we saw a rise of large corpora for second language acquisition (cf. among others, Granger 2003 and 2012, Hawkins and Buttery 2009). Beside the fact that few corpora are freely available to the research community (see however Milde and Gut 2002, Tortel 2008, Herment et al. 2012a and 2012b, and Gut » 2009), a glimpse at the « Learner corpora around the world database (http://www.uclouvain.be/en-cecl- lcworld.html) reveals that: - Most existing corpora concern English - Most existing corpora are written and not spoken - The pair of languages French-Korean is quasi-inexistent. Still, the use of large corpora allows a better evaluation of possible correlations between the learner’s L1 (first language), his grammatical competence and his proficiency level in L2 (second language) (following for example the Common European Framework of Reference for Languages). Corpus-based studies can be used to determine how some morpho-syntactic phenomena are acquired in English as a foreign language (see for example the project English Profile, Hawkins and Buttery 2009), how students’ pronunciation of L2 can be influenced by his L1 (I-PFC project, Racine et al. 2011, 2012) and so on. Depending on how the corpus is built, such corpora can also be useful for understanding learners’ opinion of the target language during education. In this paper, we present a bilingual corpus of French learners of Korean and Korean learners of French which aims at giving an answer to this missing linguistic resource by putting in parallel two languages, French and Korean. There are today numerous corpora of French as L1, be they written or spoken (the Frantext corpus (http://www.frantext.fr/), TCOF corpus (http://www.cnrtl.fr/corpus/tcof/), PFC (http://www.projet- pfc.net/) and so on). For instance, the TUFS corpus is a 1 CBFC stands for “Corpus Bilangue Français Coréen” (French Korean Bilingual Corpus). The term bilangue is to be representative example of …

Content truncated to 500 words