This course will provide an in-depth study of the MT paradigms (direct, transfer, statistical/example, and interlingual) used in state-of-the-art speech-to-speech and text-based MT systems, from computational and linguistic perspectives. Machine-learning techniques for processing human languages (morphological analysis, tagging, syntactic and semantic parsing, and language generation) will be discussed in detail. Linguistic variation across languages and its impact on computational models will be presented. Projects will involve implementing speech/text translation components, identifying their limitations and suggesting improvements.
