Bio

Kenneth Heafield is a Lecturer in the School of Informatics at the University of Edinburgh where he leads a machine translation group. He wrote the KenLM library for efficient n-gram language modeling and now works to make neural machine translation faster and higher-quality.  He holds a PhD from Carnegie Mellon's School of Computer Science and did a postdoc at Stanford.

Research interests

Kenneth combines natural language processing and systems to make and scale models for machine translation. Neural networks have improved translation quality and their ability to learn, or not learn, natural language phenomena is an open research area. However, the high computational cost of training neural network models has slowed experimentation and has forced researchers to decrease training data sizes, sometimes by three orders of magnitude. Through The Alan Turing Institute, Kenneth is collaborating with Intel to accelerate neural networks and challenge the HPC community with real-world natural language processing tasks.