Context Based Afaan Oromo Language Spell Checker For Handheld Device

Abstract

Spellchecking is a spelling check app that will carefully go through your text to scan it for any spelling errors and correct them by providing possible ranked suggestion for user to select from list and fix misspelled words. This thesis describes the design architecture, implementation and testing of a model that have been developed by a programing language Python. This spellchecker came with an integrated user friendly graphical user interface, where users can input their text, detect misspelled words and choose from a list of five candidate correction words to correct them. Users can even add words to a pre-built dictionary. Error detection is based on the dictionary look up method, bigram and trigram analysis. The data collected from the different scientifically and error free as well as trusted sources and prepare the dictionary, bigram and trigram model for error detection and correction. Two types of error happened in spelling check system to detect and correct both context aware/ real word and non-word error types. The main focus of this study is to design context based spell checker for Afan Oromo language hand held devices depends on the spelling error patterns of language based on the sequence of words in the input sentences contextually. The first types of spelling error that is non-word error candidate generation is based on dictionary lookup techniques, similarity is measured using the Levenshtein edit distance by considering Insertion, deletion, substitution and transposition of character of user input to the dictionary token and ranking top 5 probable suggestions accordingly. The second types of errors occur during spell cheek that is the real word error, for this types of error the bigram and trigram model created from the corpus and Stord based on statically/probabilistic analysis techniques was used to identify the misspelled word based on context to correct bad word according to context misspelled. To conduct experiment 1500 words were used to learn and test the model respectively. Experiment result shows that, the accuracy of 85% for spelling errors. According to gated result the accuracy of the system is 85%, this shows that the model is convenient and efficiency in order to correct misspelling Afan Oromo words both real word and nor word types of spell error occurred while user type texts to communicate.