Volume 50, Issue 1

This article provides an overview of methodological and technical issues that arise in the collection, indexing and use of spoken learner corpora, i. e. corpora containing spoken utterances of learners of a target language. After an introductory discussion of the most important special features of this type of corpus that distinguish it from written language learner corpora and spoken corpora with L1 speakers, we will go into more detail on questions of corpus design. The main part of the paper is then an overview of the methodological and technical procedures of the individual steps of collecting, indexing, providing and using spoken learner corpora. The main aim of this overview is to highlight practices that can be considered best practices according to the current state of research. Finally, we outline the challenges that still exist for this type of corpus.

Phonetic learner corpora represent a special type of spoken learner corpora by providing detailed phonetic and phonological annotation in the form of phonetic transcription as well as the segmentation and labelling of the speech signal on the levels of segments, syllables, words and sentences. This time-consuming post-processing enables a better acoustic analysis of the data and provides many options for using audio data in teaching foreign languages. It also offers good opportunities to develop tools for computer-aided pronunciation training. The possibilities and restrictions of phonetic learner corpora are illustrated with the bidirectional German–French IFCASL corpus.

This chapter introduces the multilingual Swiss learner corpus SWIKO while also exploring effects of contextual factors on teenage learner productions in instructed foreign language settings. Corpus-based language acquisition research relies on learner language data usually elicited through a range of tasks. However, task characteristics and conditions may influence the extent to which learners can demonstrate their language competences. To tackle this challenge, careful attention was devoted to the systematic variation of parameters (eight different tasks, solved orally or by writing, in the language of schooling or one or two foreign languages, or paper- or computer-based). These are first discussed in light of the overall composition of our corpus. We then focus on the German sub-corpus, reporting findings regarding text length, lexical diversity, and cross-linguistic influence. Our analyses reveal both quantitative and qualitative differences depending on task type and modality, and show large intra- as well as interindividual variation among learners. Despite the limitation of a (currently) small oral data set, these initial findings emphasize the crucial role of contextual factors in data collection and raise concerns about the generalizability and comparability of basic-level learner productions.

In this corpus-based study we explore three measurements of L2 fluency – articulation rate, filler particles, and pauses –, both within and between two registers of spontaneous dialogues spoken by Polish learners of German. The measurements are assessed both in toto (as calculated over the whole dialogue) and in parte (as calculated for specific sections). The sections are identified on a quantitative tier that divides the dialogue into four parts, and qualitatively on two linguistically-informed tiers, comprising sections based on dialogue move and task. We challenge the assessment of fluency as an average measurement over the entire dialogue, showing that a sectionwise analysis offers a better understanding of similarities and differences both within and between the two registers.

We investigate the linguistic complexity of oral classroom interactions in late primary and early secondary school across German school types. The goal is to explore whether teachers and students align in terms of their use of the academic language register. We empirically base this investigation on transcriptions of teacher and student contributions during content matter lessons on the vaporisation and condensation of water. Across school types and grade levels, we compare the extent to which teachers offer language that is adaptively rich in linguistic constructs commonly associated with academic language, such as deagentivation, nominal style, and cohesive devices. Putting this in relation to the developing academic language competence of the students, we then compare the language offered by the teachers to the use of these academic language constructs in the students’ spoken language contributions. We discuss the methodological challenges arising from analyzing oral classroom interactions and from applying automatic linguistic complexity analyses to such data.

This article discusses some methodological and conceptual issues arising from the analysis of advanced-level spoken language corpora. This discussion is based on GeWiss, a corpus of spoken academic discourse comprising L1 and L2 data of German collected in academic settings in Poland, the UK, Bulgaria, and Germany, as well as L1 data of Polish, English, and Italian. The data were collected using an ethnographic comparative approach; a range of speech events with comparable institutional and functional features in the philologies (student presentations, research papers, and oral examinations) were identified, recorded, transcribed, and tagged. In addition to POS tagging and lemmatisation, in selected sub-corpora some pragmatic features were also annotated. These include metadiscursive actions as well as references and quotations. The corpus therefore comprises ‘natural data’, derived from various contexts, of speakers with varying degrees of academic socialisation and competence in academic German. The paper discusses the various factors that have to be taken into account when analysing and evaluating the GeWiss L2 data with regard to style and register of metadiscourse as well as references and quotations in L2 student presentations. It is argued that any attempt to assess these data needs to reflect upon the concepts of ‘norm’, ‘appropriateness of usage’, (stage and context of) academic socialisation, and the context of the speech event. This also applies to the evaluation of the L1 data, which is why they cannot be used as a simple reference point for the purpose of a (mainly quantitative) ‘contrastive interlanguage analysis’ (Granger 2015). A qualitative approach needs to be used at least initially to identify and assess specific features of L2 data. This is shown using examples of metadiscursive and referencing speech actions. While there seem to be some stylistic and register features that can be regarded as inappropriate irrespective of the context of the speech events, others seem to be more context-sensitive. The paper argues that in addition to linguistic analysis and the intuitions of the L1 researcher, some kind of rating procedure would help to establish better concepts of ‘norm’ and ‘appropriateness’ of such natural corpus data originating in different academic contexts and traditions.