skip to primary navigationskip to content
 

Human Language Technologies

Providing the computational modelling of speech and language needed both to inform theoretical developments and to provide translational pathways to applications

Statistical model of speech
Image: Dr Xunying Liu
Human speech and language is so varied and complex that many aspects of it can only be investigated with large-scale computational models.

The scope for exploration is vast: we are a long way from having the deep understanding that would allow us to produce artificial systems with human-like language capability, such as would be necessary to sustain a natural conversation, to understand a newspaper article or to explain the behaviour of some complex computer algorithm. However, in the last few years, researchers in Cambridge and elsewhere have developed useful applications despite the current limitations of our knowledge. For example, systems can engage in a spoken dialogue about a restricted topic, extract useful information from collections of scientific papers and grade essays written by language learners. There is also huge potential for applying the technology which underlies these applications to new areas.

The Human Language Technologies (HLT) theme concerns the development of computational models both for theoretical understanding of speech and language and for the development of real applications. It cross-cuts the other research themes in the Cambridge Language Sciences Strategic Research Initiative, since it includes implemented approaches to language assessment, learning and change, as well as to language comprehension and production. In some cases, the HLT approach enables new ideas to be tested under realistic conditions, while in others, new theories arise from computational thinking. It also provides a pathway to practical applications. In Cambridge, research in Human Language Technologies also has strong interactions with groups whose interest is not directly language-related, including machine learning, artificial intelligence, theorem proving, mathematics and biomedical and chemical informatics.

The Cambridge Language Sciences Initiative allows HLT researchers to work in the truly interdisciplinary way which is essential for fuller understanding of human speech and language and for the development of transformative applications.

Professor Ann Copestake, Computer Laboratory

Professor Phil Woodland, Department of Engineering

People specializing in this area