Development of Models For Non-Word Error Detectionfor and Correction System for Kiswahili Language
Loading...
Files
Date
2023-11
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Kampala International University
Abstract
The development of effective spellcheckers for Kiswahili, such as the JAMBO SPELL CHECKER (2004), has marked significant progress; however, a noticeable research gap persists in the field of non-word error detection and correction systems tailored specifically for the Kiswahili language. This study addresses this gap by proposing and implementing enhanced models, SwaDetect and SwaCorrect, designed to adeptly detect and rectify non-word errors in Kiswahili. With a focus on addressing limitations in scope, speed, and accuracy prevalent in existing solutions, our research aims to pioneer a more comprehensive and efficient system for non-word error detection and correction in Kiswahili. Notably, experimental results demonstrate SwaDetect's exceptional accuracy of 99% in Kiswahili word detection, operating at a processing speed of 65 Hz (65 words per second), while SwaCorrect proficiently corrects erroneous words with an average accuracy of 82% for Edit Distance One (ED1) through ED3, maintaining an overall correction speed of 58 Hz (58 words per second). Our study encapsulates the development of models crucial for advancing the accuracy and speed of non-word error detection and correction systems dedicated to the Kiswahili language
Description
A project submitted to the faculty of information technology in fulfillment of the requirements for the award of the degree of master of information system of kampala international university