CItA: an L1 Italian Learners Corpus to Study the Development of Writing Competence

CItA: an L1 Italian Learner Corpus to Study the Development of Writing Competence Alessia Barbagli•, Pietro Lucisano•, Felice Dell’Orletta⋄, Simonetta Montemagni⋄, Giulia Venturi⋄ •Dipartimento di Psicologia dei processi di Sviluppo e socializzazione, Università di Roma “La Sapienza” alessia.barbagli@gmail.com,pietro.lucisano@uniroma1.it ⋄Istituto di Linguistica Computazionale “Antonio Zampolli” (ILC–CNR) ItaliaNLP Lab - www.italianlp.it {name.surname}@ilc.cnr.it Abstract In this paper, we present the CItA corpus (Corpus Italiano di Apprendenti L1), a collection of essays written by Italian L1 learners collected during the first and second year of lower secondary school. The corpus was built in the framework of an interdisciplinary study jointly carried out by computational linguistics and experimental pedagogists and aimed at tracking the development of written language competence over the years and students’ background information. Keywords: Italian Learner Corpus, Diachronic Evolution of Written Language Competence, Error Annotation 1. Introduction Over the last ten years, language technologies have been successfully exploited to study the development of lan- guage learning processes. A variety of different approaches based on Natural Language Processing (NLP) tools has been developed for different purposes, such as to track the syntactic development in child language (Sagae et al., 2005; Lu, 2007; Lubetich and Sagae, 2014), to measure the de- velopmental language progress using child speech patterns (Sahakian and Snyder, 2012). NLP–based approaches have been devised also to detect mild cognitive impairments us- ing measures of syntactic complexity (Roark et al., 2007) or of semantic and pragmatic atypicality (Rouhizadeh et al., 2013), and to select reading material that are appropriate for students’ reading proficiency considered a fundamental component of language competency (Schwarm and Osten- dorf, 2005; Petersen and Ostendorf, 2009). As witnessed by the increasing success of the Workshop on Innovative Use of NLP for Building Educational Applications (BEA) arrived in 2016 at its eleventh edition1, language technolo- gies have been also exploited in educational settings to design and develop educational applications such as for instance Intelligent Computer–Assisted Language Learn- ing systems (ICALL) (Granger, 2003) or Automatic Essay Scoring systems (Attali and Burstein, 2006). For all these studies and applications, the availability of electronically accessible corpora of student essays is of piv- otal importance. So far, several learner corpora have been built mainly differing at the level of typologies of collected essays (i.e. written or speech transcriptions), goals of anal- ysis (e.g. theoretical studies or development of educational applications), typologies of considered learners (e.g. learn- ers of first or second language, adults or children). A spe- cific interest has been devoted to the construction of written learners’ corpora where errors (i.e. erroneous forms written by learners) are annotated and classified; this is especially (but not only) the case of corpora of students learning a 1http://www.cs.rochester.edu/∼tetreaul/naacl-bea10.html foreign language (L2). Corpora enriched with this kind of information can offer insight into learners’ development of competencies and difficulties (Deane and Quinlan, 2010), they can be used to investigate the characteristics of in- terlanguage (Brooke and Hirst, 2012) or as reference re- sources for automatic error detection and correction tasks. The latter is the case of the …

Content truncated to 500 words