From Wikipedia, the free encyclopedia
Jump to navigation Jump to search
SkELL: Sketch Engine for Language Learning
SkELL logo.svg
Concordance English language learning – Sketch Engine
Concordance English language learning – Sketch Engine
Original author(s) Vít Baisa, Vít Suchomel
Developer(s) Lexical Computing Ltd.
Initial release November 2014; 3 years ago (2014-11)
Written in jQuery, JavaScript
Available in English, Russian, Czech
Type Language learning
License freeware

SkELL is an abbreviation of Sketch Engine for Language Learning.[1] It is a web interface for language learning. The main purpose is to help students and teachers of languages. SkELL has its own corpus[2] that was gathered so that contained texts covering everyday, standard, formal, and professional language.[1] In the corpus, there are a total of more than 60 million sentences and more than one billion words.[3]

The SkELL interface provides features such as simple search showing words in context, but the maximum of displayed lines (concordances, in fact) is 40. However, the frequency of searched query is located below the search box and expressed with the number hits per million. The second function is word sketch which enables showing collocates for a given word or words. The last one is named as similar words. It visualises similar words to searched word in a word cloud.

The tool has been available also for the Russian language (since 2015)[4] and the Czech language (since 2017).[5]


SkELL offers three types of searches.

  • Examples – searching for words and phrases and their all derived forms
  • Word sketch – a simplified version of the original word sketch page
  • Similar words – based on the Distributional thesaurus in Sketch Engine, there are not necessarily synonyms


The corpus for English SkELL consists of English Wikipedia (special sorted out 130,000 articles), English collection of Project Gutenberg, a subset from the web corpus enTenTen14,[6] the whole British National Corpus, and free new sources.[1]

Processing the data[edit]

After gathering and pre-cleaning (all structures have removed except sentences) data has run through processing pipe: normalization, tokenization, TreeTagger for English, and deduplication. The further process was a compilation of the corpus using manatee indexing library. In the end, all sentences were scored with the GDEX[7] tool.[1]


  1. ^ a b c d Baisa, Vít; Suchomel, Vít (2014). "SkELL:Web Interface for English Language Learning" (PDF). Eighth Workshop on Recent Advances in Slavonic Natural Language Processing. NLP Consulting: 63–70. 
  2. ^ Thomas, James (14 June 2015). "Discovering English with SketchEngine – James Thomas interview". EFL NOTES. Retrieved 21 December 2015. 
  3. ^ "SkELL". Sketch Engine. Lexical Computing Ltd. Retrieved 21 December 2015. 
  4. ^ Valentina, A., Vitalevna, B. O., Малолетняя, А. П., Olga, K., & Vit, B. (2016). RuSkELL: Online Language Learning Tool for Russian Language. In Proceedings of the XVII EURALEX International Congress. Lexicography and Linguistic Diversity (6–10 September, 2016) (pp. 292-300). Ivane Javakhishvili Tbilisi State University.
  5. ^ Cukr, Michal (2017). Český korpus příkladových vět (Czech corpus of example sentences) (Master's thesis thesis) (in Czech). Brno: Masaryk University, Faculty of Arts. Retrieved 2017-06-22. 
  6. ^ Jakubíček, Miloš.; Kilgarriff, Adam; Kovář, Vojtěch; Rychlý, Pavel; Suchomel, Vít (July 2013). "The tenten corpus family". Seventh International Corpus Linguistics Conference CL. Lancaster University: 125–127. 
  7. ^ Kilgarriff, A.; Husák, M.; McAdam, K.; Rundell, M.; Rychlý, P. (July 2008). "GDEX: Automatically finding good dictionary examples in a corpus". Proceedings of the XIII EURALEX International Congress. EURALEX: 425–432. 

External links[edit]