This document provides a summary of statistical machine translation (SMT) in a series of slides. It introduces key concepts in SMT, including modeling translation as a probability, decoding source sentences to find the most probable translation, training probability models from bilingual corpora, using log-linear models to combine multiple feature functions, and optimizing feature weights to maximize translation performance on a tuning set as measured by metrics like BLEU. The typical SMT development process of training, tuning, and evaluation is also outlined.