Spoken learner corpora : diversity and comparability
Résumé
This presentation discusses the methodological issues involved in finding a satisfactory compromise between common conventions of transcription and annotation in spoken learner corpora and the specific requirements of compiling corpora in different settings-primary, secondary and higher education-different languages and different modes of production.
This presentation will look at the theoretical and methodological issues involved in compiling spoken learner corpora across three different institutional learning contexts (primary school, secondary school, university), in varying languages and different conditions of L2 production. Specifically, the characteristics of the corpora are as follows : A corpus of spontaneous monologue production (university students, L2 English, French and Italian) A longitudinal corpus of monologue and peer interaction (university students, L2 French, various L1s) A corpus of classroom interaction (primary school, L2 English) A mixed corpus of spoken and written production (secondary school CLIL class, L2 English) The research projects within which these corpora have been compiled focus on different aspects of L2 learning/teaching – fluency, communication strategies, patterns of classroom interaction, relations between input and output – with corresponding differences in design and transcription/annotation choices. At
the same time, each corpus is potentially open to other types of analysis, and it is therefore desirable that a certain number of transcription and annotation procedures should be held fairly constant. The presentation will discuss the methodological issues involved in reaching a reasonable compromise between specific requirements and common conventions, and the advantages and limitations of the tools used in the compilation and analysis of the corpora : CLAN, ELAN and EXMARaLDA.