Towards a gold standard for Swedish CEFR-based ICALL
Elena Volodina, Dijana Pijetlovic, Ildikó Pilán, Sofie Johansson Kokkinakis · 2013
In qualitative projects on ICALL (Intelligent Computer-Assisted Language Learning), research and development always go hand in hand: development both depends upon the research results and dictates the research agenda. Likewise, in the development of the Swedish ICALL platform Larka, the practical issues of development have dictated its research agenda. With NLP approaches, sooner or later, the necessity for reliable training data becomes unavoidable. At the moment Larka's research agenda cannot be addressed without access to reliable training data, so-called “gold standard”. This paper gives an overview of the current state of the Swedish ICALL platform development and related research agenda, and describes the first attempts to collect the reference corpus (“gold standard”) coming from course books used in CEFR-based language teaching.