MODELING AND IMPLEMENTING AMHARIC NON- STANDARD WORDS SPELLING CHECKER AND CORRECTOR

Abstract

Amharic is a language that is spoken by millions of people in Ethiopia and Ethiopian living internationally. It is a widely used language for creating documents for communication purposes. However, since no spelling checker computer program detects and corrects for non-standard Amharic language, spelling errors are becoming common and interfere with communication. A spelling checker is a computer program that detects and often corrects misspelled words in a text document. In response to this problem, the researcher sets developing and implementing nonstandard error Amharic language spell checker and corrector model. To achieve this objective the researcher uses design science research methodology. This research is aimed at modeling and designing Amharic non-standard words spell checker and corrector, and hence research of this nature is best addressed through design science research methodology. The researcher uses dictionary lookup for error detection technique and minimum edit distance as error correction technique. While dictionary lookup detects misspelled words sequence matcher provides spelling suggestions and the list of candidate spellings. The research also employs tools for compiling the python code and storing the corpus. It also uses tools for text processing and for developing a graphical user interface. To demonstrate the validity of the non-standard Amharic words spelling checker and corrector model and to measure its accuracy, precision and recall, confusion matrix have been used as measuring matrix. As a result the model precision, accuracy and recall has 0.94, 0.93 and 0.87 respectively. A questionnaire is also prepared to measure the usability of the prototype on basis of ISO 9241 usability engineering standards and distributed to respondents who are familiar with Amharic Writings. Based on this, the research finds out that the accuracy of the model designed for non-standard words Amharic language is 92%. This clearly shows that the model is effective in checking and correcting words written in Amharic.