Sampson G. Evolutionary Language Understanding

London, New Delhi, New York: Bloomsbury Academic, Bloomsbury Publishing Plc, 2015. — XIV, 210 p.The author provides a context for understanding the differences in assumptions about the nature of language and cognition. Evolutionary language understanding is beginning to look like a technology with a bright future. The discipline within which the work described here is situated has come to be known, perhaps more frequently in Europe than in North America, as language engineering. In the past, computational approaches to 'natural language' - human languages such as English, German, Japanese - have often been seen as an exception. Language is obviously a leading manifestation of the distinctively intellectual nature of our species, and researchers have seen the development of computational models of natural language as a technique for studying the nature of human cognition. The programme of work reported in the following chapters has not been conducted in that spirit. It may be located at the 'researchy', intellectually innovative end of the language engineering spectrum, but it is intended as an engineering exercise, aiming to develop reliable systems for executing practically useful functions related to natural language. In so far as humans can be said to execute the same functions, there is no suggestion that they do it in the same way.The idea that studying the behaviour of the glorified calculating machines we call computers can explain the cognitive life of creatures endowed with the divine spark of creativity tends to be most persuasive to people whose direct experience of computers is fairly limited. At an early stage of my career I imagined it might be true myself. Some decades later, I no longer believe the two topics have much to do with one another; but this does not matter - it is clear that computers can be successfully harnessed to execute various economically useful language processing tasks, without our needing to enquire how people process language.Prologue
The Origin of the Task
Simulating Evolution
An Early System
The Current System: Goals and Standards
The Current System: Front End and Parsing Algorithms
The Current System: Language Model
Parallel Tree Optimization
Epilogue