Design and development of a bangla spell checker

Abstract

This research work has been aimed at designing and developing a BangIa Spell Checker. Use of different forms of same letter with different set of rules in BangIa, makes the straightforward development of the BangIa Speller very tedious. To search the alternative(s) of the miss-spelled word several ways have been considered. For suggestion, words having similar sound near to the miss-spelled word have been searched based on sound code of BangIa words. To generate sound code, sound of each letter of the BangIa alphabet has been analyzed and grouped according to their phonetic similarities and based on the effect upon the human vocal tract. The composite consonant has been 'decomposed and grouped. By varying the matching part of sound code the nuniber of suggestion is controlled. Coding system has been developed for detecting and suggestion searching for the most natural error - the replacement error. Also, the miss spelling due to keyboard adjacency error based replacement error has been prioritized with possible suggestion order at the top. With advent of coding, controlling over suggestion pattern and number becomes very effective. Same coding system is also used for deletion error, Insertion error and swap error based suggestion searching, effectively. The complexity of non-coded system is removed with the coding system. Users sometimes wrongly choose similar sounding but different letters out of confusing which lead to miss spelling. This kind of minor error detection has been executed with topmost priority. Coding system is also effectively introduced in this case with different levels of controlling. In the coding system, impact of vowel and consonant has been differentiated in case of sound code generation. Non-principal form and principal form of letter has been given very lower and very higher value respectively for other error based searching. To develop the coding system non-weighted component of a word has also been separated as a prerequisite. This simplified or open formai is an added parameter for the suggestion searching. Along with development of the Speller the development of the interface overcoming possible difficulties has been shown. Ultimately, the code based combined searching strategies has formed a strong platform for the BangIa Spell Checker and the side-by-side development of the Speller interface helps reaching the completeness.