EXPLORE PUBLICATIONS BY COUNTRIES

Diacritic Restoration, Northern Sotho.

Diacritic markers are usually not inserted in text for convenience as users type text. However, text without diacritic markers could affect the quality of its analysis as it may affect how it is pronounced and its meaning among others. The number of diacritics and the impact of not inserting them vary from language to language. The processes of restoring diacritics in the text can be looked at as language-dependent and language-independent and also as word-based or syllable based. Northern Sotho language uses two diacritic markers to indicate pronunciation and also distinguish between homographs in some cases. Very little research has been done on diacritics restoration in the Northern Sotho language. In this paper, we show that morphological word transformations are consistent in how they insert or do not insert diacritics in derived words. We focus on the caron diacritic marker.An input word is reduced to its root form by a morphological analyzer. The accented form of the root word is retrieved from the diacritic dictionary. This word, together with morphological rules is used to determine the diacritics of the input word. The implemented tool gave a recall performance of 86% on test data. Most errors were due to failures in the morphological analysis of the input word.