Improving a Spelling Checker for Afrikaans Menno van Zaanen and Gerhard van Huyssteen University of Amsterdam, University of Tilburg Potchefstroom University for Christian Higher Education (CHE) Abstract In this paper we describe the development of an improved spelling checker for Afrikaans. We compare two currently available spelling checkers and discuss their shortcomings. The existing applications are restricted in their suggestion capabilities, as well as their preci- sion and recall, mainly because they cannot treat morphologically complex words correctly. Here, we will mainly focus on improvements in precision and recall. The general architecture of the existing spelling checker is discussed and several im- provements are implemented. We describe an improved lookup phase and a newly added morphological analysis phase. The morphological analysis poses some problems which are also treated. Finally, some remaining problems are mentioned. 1 Introduction Although Natural Language Processing and Computational Linguistics have been of extensive research interest for many years in the USA, Europe, and other parts of the world, these fields are by and large unexplored territory in South Africa. Except for some sporadic research efforts by individuals, no Human Language Technology (HLT) projects existed for many years. Only recently an awareness of the importance of HLT for the South African community has started to bloom on both academic and governmental level, to such an extent that, in 2002, the minister of the Department of Arts and Culture convened an advisory committee to steer the development of HLT in South Africa. Since 2001, the Potchefstroom University prioritized these fields as strategically important research topics, and established a research focus area called “Language and Technology”. The university invested extensively in the enhancement of personnel and other resources to develop human language technologies for some of the South African languages, including the establishment of the first complete graduate study pro- gram in Computational Linguistics in South Africa, the setting up of a dedicated language technology laboratory, and the acquisition of text and speech corpora for Afrikaans, South African English, and Setswana. The first major HLT project to be funded by the university was a project to develop a spelling checker for Afrikaans, based on morphological analysis. As the IT department at the Potchefstroom University had already developed a spelling checker for Afrikaans (the PUK/Microsoft Spellchecker, here called PUKspell), the development of an enhanced spelling checker was considered an “easy” project to get hands-on experience in the fields of Natural Language Processing and Com- putational Linguistics, and at the same time to develop reusable enabling tech- nologies for Afrikaans. Over and above these educational and research purposes, 1