PROTOTYPING NLP NON-WORD DETECTION SYSTEM FOR DINKA USING DICTIONARY LOOKUP APPROACH

Spellcheckers are computer software used for non-word or real word error detection. The Dinka text
editors have been developed, however, no one has developed their spellcheckers. The research entitled
Prototyping NLP Non-Word Detection System for Dinka Using Dictionary Lookup Approach was a
solution to Dinka spellchecking. The study objectives were: requirements gathering and analysis. The
computer keyboard was customized to accept the Dinka characters. Dinka lexicon was created with 6,976
words. The prototype was implemented using java programming language and dictionary lookup approach
was used for non-word detection. The accuracy of detection (detecting real words and non-words) gave
98.10%, and the accuracy of non-word detection (detection of non-words only) was 91.36%. The True
Positive Rate (TPR) was 99.10% and the True Negative Rate (TNR) was 91.36 %. The speed of non-word
detection which was found as1, 044 Hz was slow.