Language Learning Round Table: Corpora as resource for second language research and teaching. Opportunities and challenges (part 2)

The main use of corpora in language instruction is to provide various types of word lists. The intuitive word lists of bygone days are now replaced by objective lists extracted from corpora by computer software (such as concordancers) and deployed by researchers, teachers, and learners via computer software (such as lexical profilers). Corpus lists still involve judgments, however, mainly to do with the way their words will be assembled into groups. Early computational lists were mainly word family based, but in the past 10 years the word family has come under criticism and the lemma proposed as its replacement. (Families are the inflected plus derived forms of a head word; lemmas are inflected forms only.) Despite acceptance in principle of the lemma as the basis of pedagogical word lists by many researchers, however, there is as yet no fully worked out example of what a comprehensive lemma list would look like or how it would function. My talk will provide such an example. Based on my work with word lists on the Lextutor website, I will outline the main uses language practitioners make of such lists and show how a pure lemma approach would not work for any of them. I will conclude by proposing a word unit and technology for providing it which is neither family nor lemma but a principled synthesis of both.

The proficiency of foreign language learners is often evaluated based on their productions. However, the complex interplay between task-, learner- and rater-related variables continues to challenge teachers and researchers alike. As such, open questions remain in regards to the extent to which individual texts can reliably and adequately represent the linguistic competence of a learner, and which text measures are associated with the human rating of a learner production's quality.
We aim to address these questions for lower-level proficiency learners based on SWIKO, a multilingual learner corpus currently being developed at the Institute of Multilingualism, Fribourg (CH). The corpus comprises of oral and written productions by Swiss lower secondary school students both in their language of schooling and foreign languages. Participants completed eight communicative tasks which systematically vary by genre, topic, and structuredness, therefore allowing to disentangle the effects these task features may exert on the productions. Following the collection and corpus linguistic processing of the data, prospective foreign language teachers rated a part of the productions based on the scales of the Common European Framework of Reference (CEFR, adapted and partly validated for young learners). Focusing on the written texts in German and French as foreign languages, we present preliminary findings on the connection between (corpus-)linguistic measures and human ratings, as well as the role of different task features in this relationship.