Machine Learning for Efficient Natural-Language Processing
Much of computational linguistics in the past thirty years assumed a ready supply of general and linguistic knowledge, and limitless computational resources to use it in understanding and producing language. However, accurate knowledge is hard to acquire and computational power is limited. Over the last ten years, inspired in part by advances in speech recognition, computational linguists have been investigating alternative approaches that take advantage of the statistical regularities in large text collections to automatically acquire efficient approximate language processing algorithms. Such machine-learning techniques have achieved remarkable successes in tasks such as document classification, part-of-speech tagging, named-entity recognition and classification, and even parsing and machine translation.