Abstract:
We present a word based Bangla spelling checker which improves the quality of suggestions with the help of the previous and next words of the misspelled words in a document. A spell checker, as we know, is a tool used for checking the spelling errors. Also, it corrects those errors in the text or a document. Development of any application for Bangla language is relatively complicated due to the complexities of the Bangla character set. Bangla alphabet consists of nearly 160 complex shaped compoundcharacterclassesinadditionto50basiccharacterclasses. Obviously,developing a spell checker application for Bangla language raises many new difficulties which do not have to be dealt with in case of a Latin based texts such as English. Most of the currently available Bangla spell checkers are based on correcting errors that has been committed on a character level. The complex rules for Bangla spelling and the complexitiesofBanglacharactersetdemanddifferenterrordetection&correctionmethods from those used for other languages. Additionally, the presence of similarly shaped characters, compound characters and the inflectional nature of the laguage present a significant challenge in producing suggestions for a misspelled word when employing the traditional methods. Considering the intricacies of the problem we have proposed, in this paper, the development of a word based spell checker for Bangla language. Our researchisaimedatthecorrectionofmisspelledwordsbyconsideringtheirneighbouring words. For this purpose we have built a lexicon of unique structure. Based on this specially built lexicon the proposed system attempts to predict the misspelled words in an input text file. This paper also shows the performance and evaluation of our proposed solution. Finally, we conclude by describing the limitations of the system with possible future improvements.
Description:
We are thankful to Almighty Allah for his blessings for the successful completion of our thesis. Ourheartiestgratitude,profoundindebtednessanddeeprespectgotooursupervisor, Dr. Hasan Sarwar, Professor and Head of the Department, CSE, United International University(UIU), for his constant supervision, affectionate guidance and great encouragement and motivation. His keen interest on the topic and valuable advices throughout the study was of great help in completing thesis.
We are especially grateful to the Department of Computer Science and Engineering (CSE) of Military Institute of Science and Technology (MIST) for providing their all out support during the thesis work. Also, we want to express our deepest gratitude to the reviewers, Lecturer Jahidul Arafat and Lecturer Wali Mohammad Abdullah for their valuable inputs which helped us in revising the initial draft and preparing the final paper.
Finally, we would like to thank our families and our course mates for their appreciable assistance, patience and suggestions during the course of our thesis.
