Spelling Error Pattern Analysis of Punjabi Typed Text

Abstract

Error pattern analysis of a language is useful in language related technology development, such as Spell Checker and Corrector, Optical Character Recognition, Machine Translation, Natural Language Interfaces etc. Error pattern analysis includes analysis of various types of errors (insertion, deletion, transposition, substitution, run-on, split word error) positional analysis, word length effects, phonetic errors, first position error analysis, keyboard effects etc. Though considerable work has been done in the area for English and related languages, the Indian Language scenario presents a relatively more complex and uphill task. In this thesis, I have presented a statistical error analysis for Punjabi, the world’s 14th most widely spoken language. For this purpose I have collected about 20000 misspelled words generated by typists. The application of the error analysis in designing the suggestion list for a Punjabi spell checker is also discussed.