ABSTRACT

Th e standard way for communicating with a computer today is awkward. We have to laboriously move a mouse around a screen or type in commands. Wouldn’t it be nice if we could just talk to a computer or robot and tell it what to do? Spoken language is the natural and effi cient way that people communicate with one another (Figure 6.1). If we could get machines to comprehend and produce human language, we could interact with them more easily. Imagine having a conversation with a housecleaning robot. You could tell it to clean the bathroom but not the kitchen and to wait until aft er dinner before washing the dishes. If the robot was confused about these orders, it could then ask and you would respond to clarify its confusion. In this chapter we examine some of the amazing progress that has been made in machine language. We fi rst describe what language is and then describe its human development and neural foundations. Th e bulk of the chapter is devoted to explaining the diff erent information processing stages machines go through in order to comprehend and produce language and to engage in conversation. Although a number of hurdles remain, we can see from this work that computers are capable of reproducing much of human linguistic ability. You may be surprised to discover that there are already artifi cial conversational agents that can engage in complex typed or spoken dialogues with people. From a functional perspective, these agents understand and use language in a way that is at times indistinguishable from humans.