Towards a Corpus-based Approach to Modelling Language Production of Foreign Language Learners in Communicative Contexts

Proceedings of Recent Advances in Natural Language Processing, pages 557–561, Hissar, Bulgaria, 12-14 September 2011. Towards a Corpus-based Approach to Modelling Language Production of Foreign Language Learners in Communicative Contexts Voula Gotsoulia Research Center for English Language Faculty of English Studies National and Kapodistrian University of Athens vgotsoulial@enl.uoa.gr Bessie Dendrinos Department of Language and Linguistics Faculty of English Studies National and Kapodistrian University of Athens vdendrin@enl.uoa.gr Abstract This paper discusses linguistic annotation issues, essential to a corpus-based ap- proach to modelling the language use of foreign language learners in various con- texts. We focus on learners of English and describe the corpora we use as well as the linguistic approach underlying their development. We present a scheme for de- scribing grammatical choices and meaning components expressed in texts produced by learners. Our goal is to model the asso- ciations of corpus-attested linguistic pat- terns with their contexts, at different levels of language proficiency. 1 Introduction Learning a foreign language is a complex process involving mastering a range of elements of a non- trivial system of communication and being able to use them appropriately in different social con- texts. In a related vein, assessing a learner’s abil- ity to use language (i.e. his/her linguistic com- petence) is a significantly complicated task, re- quiring well-defined criteria for describing the in- stantiations of the system of language in socially meaningful ways. The Common European Frame- work of Reference for Languages (CEFR) has at- tempted to provide an objective basis for the ex- plicit description of language proficiency across Europe, aimed to promote the transparency of lan- guage courses and ‘the mutual recognition of qual- ifications gained in different contexts’. CEFR distinguishes among several types of language-related communicative competences (i.e. lexical, grammatical, semantic, phonological, orthographic, orthoepic, sociolinguistic, prag- matic) and gives illustrative descriptors for each of these competences across the six-level scale of language proficiency established by the Council of Europe.1 These descriptors are formulated in a very general way. In practice, incorporating their insight into concrete models of language learning and assessment is an open issue.2 In this paper, we address the foundations of a corpus-based approach to modelling the learn- ers’ production of language in relation to partic- ular communicative contexts. Such a model can be used to support reliable assessment of language performance across proficiency levels, as well as test and materials development. We focus on En- glish as a Foreign Language (EFL) and, more pre- cisely, on the use of grammatical resources for the production of written texts. In section 2, we describe the EFL learner cor- pora we use. Section 3 presents the linguis- tic framework we essentially draw upon and dis- cusses methodological issues related to the repre- sentation of the range of grammatical resources employed by learners when producing written texts. Finally, in section 4 we specify the precise goals that we intend to pursue in the immediate future. 2 EFL Learner Corpora As a basis for our study, we use the EFL learner corpora available from …

Content truncated to 500 words