A Language Model for Spell Checking of Educational Texts in Kurdish (Sorani)
Authors
Proceedings of the 1st Annual Meeting of the ELRA/ISCA Special Interest Group on Under-Resourced Languages
pp. 189-198
June 24, 2022
Abstract
A contextual Sorani Kurdish spelling-correction study using a trigram language model with Stupid Backoff and edit-distance candidates. The authors report 88.54% overall accuracy and 43.33% F1, with substantially different performance across the educational-text evaluation subsets.
Keywords
Spell Checking
Sorani Kurdish
Language Model
Trigram
Educational Text
Related Datasets
Kurdish Textbooks Corpus (KTC)
2019
31 textbooks; 693,800 tokens; 110,297 types
XML, Text
A Sorani corpus created from Kurdish K-12 textbooks and organized by educational subject.
KTC-Language-Model
Date not verified
94,188 unigrams; 372,903 bigrams; 521,797 trigrams
XML, Excel
The released Sorani language-model data underlying the 2022 contextual spell-checking study. The paper reports 94,188 unigram, 372,903 bigram and 521,797 trigram types. This is a language-model …
Citation
Abdulrahman, R., & Hassani, H. (2022). A Language Model for Spell Checking of Educational Texts in Kurdish (Sorani). SIGUL, 189-198.