Skip to main content
Språkbanken Text is a part of Språkbanken.

SweLL-pilot

Standard reference Information

Elena Volodina, Ildikó Pilán, Ingegerd Enström, Lorena Llozhi, Peter Lundkvist, Gunlög Sundberg, Monica Sandell (2016): SweLL on the rise: Swedish Learner Language corpus for European Reference Level studies, in Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC 2016), May 23-28, 2016, Portorož, Slovenia BibTeX

Data citation Information

Elena Volodina, Ildikó Pilán, Ingegerd Enström, Lorena Llozhi, Peter Lundkvist, Gunlög Sundberg, & Monica Sandell (2016). SweLL-pilot (updated: 2016-01-01). [Data set]. Språkbanken Text. https://doi.org/10.23695/d3bg-8g24
BibTeX Additional ways to cite the dataset.
Essays written by adult learners of Swedish, manually labeled with the CEFR levels (a European scale of language proficiency levels within language learning). Collection period 2006-2015.

SweLL-pilot corpus is a corpus of essays written by adult learners of Swedish. It was collected during the period of 2006-2015, and contains 502 essays that have been anonymized and graded with CEFR labels.

There are three subcorpora in the SweLL-pilot collection:

  • SpIn - 256 essays collected from Language Introduction course (mid-term exams) for newly arrived refugees. Some of the students are recurrent.
  • Sw1203 - 141 essays collected from university students in exam setting, most of who wrote three essays each
  • TISUS - 105 essays written as a part of a Test In Swedish for University Studies. All essays are on the same topic “Stress” and of argumentative genre

Links

Annotation

Essays are manually graded using the CEFR scale by human teachers. In addition, the essays are also linguistically annotated (POS tagging, lemmatization, dependency annotation) using Sparv.

Caveats

Data collection is limited to a small geographical area and a short period of time. Although several language backgrounds are represented, the corpus is very unbalanced in this sense and as a consequence not well suited for native language identification tasks.The corpus consists of three subcorpora, each coming from a different source, with different proficiency levels, and with different colleciton periods. While the three subcorpora can be used simultaneously, care should be taken into account to ensure that artifacts from these differences do not leak into models.

Intended uses

Automated grading using the CEFR scale, anonymization, second language acquisiton studies

References

Accessible through

Access Platform Licence
CLARIN-ID, -PRIV, -NORED, -BY (https://www.kielipankki.fi/support/clarin-eula/#res)

Type

  • Corpus
  • Collection

Language

Swedish

Size

Resources: 3

Keywords

  • L2
  • language learning
  • essays

Creators

  • Elena Volodina
  • Ildikó Pilán
  • Ingegerd Enström
  • Lorena Llozhi
  • Peter Lundkvist
  • Gunlög Sundberg
  • Monica Sandell

Updated

2016-01-01

Contact

sb-info@svenska.gu.se