Greek Learner Corpus Gateway

The Greek Learner Corpus (GLC) is the largest online, freely available learner corpus of Greek as a second / foreign language (L2/FL). It brings together two error-annotated subcorpora (GLC I and GLC II) of written and spoken learner productions, each accompanied by a rich set of metadata relevant to L2/FL teaching and acquisition. GLC II was compiled within the research project Latent Aspects in L2 Acquisition (LAL2A) during 2019-2022, funded by the Hellenic Foundation for Research and Innovation (H.F.R.I./EL.ID.EK., project no. 3161), at the Aristotle University of Thessaloniki.

GLC v.1 (the second tab on the left) is a collection of about 450 written productions (~33,000 words) by adolescent learners of Greek, compiled at the Aristotle University of Thessaloniki under the supervision of Alexandros Tantos and Despina Papadopoulou (Tantos & Papadopoulou, 2014). It draws on the language-proficiency assessment tests ‘Let’s speak Greek I, II, III’, taken in the Reception Classes of Greek schools between 2010 and 2014, and was annotated with the UAM Corpus Tool. Its error-annotation scheme formed the basis for GLC v.2. GLC v.1 is also accessible through the Greek node of CLARIN, CLARIN-EL.

GLC v.2 (GLC II, the third tab) is the larger and more recent corpus. In its open release, compiled during 2019-2022, it comprises 1,101 written texts (197,713 word tokens) and 318 spoken interviews (~156,005 tokens), about 422,360 tokens in total, produced by adult learners from 36 first-language backgrounds (most frequently Russian, Arabic, Spanish, Serbian and Turkish) and spanning the CEFR levels A1–C2. It is a cross-sectional corpus, offering a snapshot of learner performance at beginner, intermediate and advanced levels. Productions were contributed by a number of institutes in Greece and abroad: the School of Modern Greek Language of the Aristotle University of Thessaloniki, the Greek-language department for women immigrants and refugees at the Lyceum Club of Greek Women (Athens), the private language school Peek at Greek (Thessaloniki), the Sismanoglio Foundation in Istanbul, and the branches of the Hellenic Foundation for Culture in Belgrade, Odessa, Florence and Alexandria.

The written productions were elicited through a Google Form organised in three sections (project description and informed consent, learner profile, and task profile), across three genres (description, narration and argumentation), with an upper time limit of one hour. Besides the digital submissions, several hand-written texts were collected and digitised following a well-defined protocol that ensured a standardised encoding of features such as eroded letters, deletions, segmentation and paragraph indicators.

Teachers and philologists with students learning Greek as a second / foreign language who would like to contribute their own students' productions are welcome to get in touch with us.

The spoken subcorpus was collected through recorded online (Zoom) interviews of 6–9 minutes on the same topics; half of the recordings have already been transcribed following the conventions of Conversation Analysis, and the remainder will be added to the Gateway progressively.

GLC v.2 is also accompanied by a comparable control subcorpus of written productions by L1-monolingual Greek speakers (~66,645 tokens), matched to the learners for age, sex and educational background and based on the same topics. This alignment enables direct comparison between learners’ and native speakers’ productions.

The corpus carries a multi-layered, stand-off grammatical error annotation, organised hierarchically as Error Domain → Error Category → Error Cluster. Four morphosyntactic domains are covered: Agreement, Gender Assignment, Aspect and Voice. For instance, Agreement is analysed into Gender (masculine / feminine / neuter), Number (singular / plural) and Case (nominative / accusative / genitive / vocative). The annotation was carried out by six trained linguists working in pairs, reaching an inter-annotator agreement of around 90%, and is documented in an annotation manual that supports consistency and reuse across future expansions of the corpus.

This web application, the Greek Learner Corpus Gateway, is the single interface to both GLC I and GLC II. It provides descriptive statistics for each corpus, error-profile plots at corpus and text level, the error annotations and metadata in searchable tabular form, advanced search and download across any combination of linguistic and extralinguistic variables, and additional dashboards, learner-analytics, concordance and part-of-speech views.

Beyond dissemination, the GLC data and its annotations are hosted on the annoteca platform, a dedicated annotation platform, currently under development, that supports collaborative corpus annotation workflows: multi-annotator annotation, inter-annotator-agreement measurement, and curation towards a gold standard. There, in the near future, the GLC annotation will be able to be maintained, reviewed and extended in a reproducible, multi-user environment, while the corpus serves as the foundation for AI-assisted annotation of new L2-Greek data.

GLC v.2 follows a clear-cut design (see the figure below) that lets researchers quickly judge whether the corpus meets their interests.