I am a Research Staff Member in the Neural Machine Translation (NMT) Group at the IBM T.J. Watson Research Center, and I have been working on algorithms for statistical or neural machine translation for more than 25 years.
During my time as a Ph.D. student and in the following years, my research focused on search algorithms for SMT (in particular dynamic-programming (DP) based algorithms) for word-based and phrase-based machine translation models. Here, I proposed a simple, sequential, block-based segmentation model for phrase-based SMT. That work has resulted in a first successful experiment with a linear model which relies entirely on binary feature functions. In addition, I have been working on beam search algorithms for extracting parallel sentence pairs from comparable data.
More recently, I have been working on neural machine translation (NMT). Here, I have been working on improving the core NMT engine, e.g. implementing batch decoding and improving the NMT system by data augmentation via back translation and cross-lingual sentence-level matching based on multilingual sentence embeddings. In addition, I have been working on automating the end-to-end build process for an NMT system: it can be automatically deployed in a cloud-based environment starting from just the unprocessed parallel data resources.