Compiling a Task-Based Corpus for the Analysis of Learner Language in Context

Corpora in linguistics and computational linguistics have traditionally been assembled from data sources such as newspaper texts, books and, more recently, the web. While these sources provide large quantities of language data, typically very little or nothing is known about the context under which the text has been produced. The only information an analysis can refer to is the text itself, e.g., when a sentence is analyzed using the preceding sentences for disambiguation. However, language is always produced in a concrete extra-linguistic context. This contextual setting includes world knowledge and situational knowledge, i.e., the aspects of world knowledge which are relevant to interpret the given text and the concrete task and situation that the language was produced for.