Thesis

My PhD thesis, entitled "Domain adaptation for neural machine translation", is available here. It was completed at Cambridge University Engineering Department in 2021. I was supervised by Prof. Bill Byrne.

The thesis explores techniques for adapting neural machine translation systems to new domains of language. I pay particular attention to techniques that allow adaptation to examples from multiple or unknown domains, and to mitigating negative side-effects of adaptation such as over-fitting, exposure bias, and catastrophic forgetting of previous examples.