Machine Learning of Natural Language

In the 1960s Chomsky decided that language couldn't be learned due to 'Poverty of the Stimulus' based on certain empirical and theoretical evidence, pushing major questions around the origins and nature of language back to biology and evolutionary theory – and then ignoring them. In the 1970s we met because we both share the same views of language, that it had to be learned, –and that Chomsky had too small a view of language.

Language is a living organism, produced by neural mechanisms relating in large numbers as a society, as we learn about our language, world and culture in the context of human society. Language exists between minds, as a way of communicating between them, not as an autonomous process; between our senses and experiences, linking past, present and future – not as a stagnant artefact.

The logical 'rules' seem to us an epiphe­nomena of the neural mechanism, rather than an essential component in language, or a component of behaviour that is essentially linguistic. This view of language has been advocated by an increasing number of workers, as the view that language is simply a collection of logical rules has had less and less success.

David Powers is a teacher of computer science and increasingly a student of cognitive science due to being disillusioned with an approach to Artificial Intelligence (AI) that ignores what we know of humans, our intelligence and how we learn. In the late 1970s he considered the work of Terry Winograd, Roger Schank and others, and decided that a totally different approach to AI and Natural Language was needed - one based on understanding the way babies learn to understand and communicate within their world.

Christopher Turk, like many workers who have come into the field of AI was originally trained in literature. but moved into linguistics, and then into computational linguistics. In the early 1980s he took a sabbatical in Roger Shank's AI project in the Computer Science Department at Yale University, becoming increasingly uneasy at the view of language which was used at Yale.