This book describes a complete translation system for spontaneously spoken language, constructed using the incremental paradigm. It starts by presenting the theoretical and algorithmic basis necessary to cope with the complex endeavour of translating speech incrementally and in parallel. In particular, graph-theoretic foundations of natural language processing and feature-based descriptions of linguistic objects are covered. A thorough description of the system and its performance follows. The author covers syntactic and semantic processing as well as transfer and syntactic generation. Thus the book can also be used as a broad-coverage introduction to the field of speech translation. This book is essential reading for researchers and students working in the field of speech translation. It is also intended as a research tool for those interested in the architecture of general natural language processing systems.