Research on computational semantics using deep learning models.

Teaching DATASCI W266 Natural Language Processing with Deep Learning.

Better machine translation system combination through positive diversity training (PDT). Training machine translation decoding models with very large discriminative feature sets. Semantic textual similarity (STS). Improved modeling of machine translation argument structure.

Machine learning for phrase-based machine translation systems, including creating the state-of-the-art Phrasal machine translation package.