Using Multiple-Sequence Alignment and Machine Learning Techniques to Improve Machine Translation

As Internet access spreads to all nations of the globe, it is becoming increasingly important to be able to access information in a language other than that in which it was written. A survey of the free machine translation (MT) systems available on the Internet yields unsatisfying results; MT systems simply have not been able to improve fast enough to satisfy the demand for reliable, accurate translation on the level that a competent human translator could produce. The present work endeavors to begin with what machine translation ends with, to work from the admittedly flawed output of these systems and, through simple techniques, to attempt to create a superior machine translation. We shall discuss two approaches for doing so: (1) by making use of multiple machine translation systems and a biological sequencing algorithm known as multiplesequence alignment, and (2) in creating an n-gram replacer for domain-restricted input texts through the application of machine learning techniques.