2.0 Neural Networks and Natural Language

Since the neural network movement became popular again in the 1980's, connectionist models have been used to analyze various aspects of natural language. Sejnowski & Rosenberg's (1986) NETtalk project, for example, drew a lot of attention. Similarly, the work by Rumelhart & McClelland (1986) on the acquisition of the past tense in English has inspired a whole series of other models. In this chapter, I will only concern myself with a few of all these models: four networks from four different areas -- computer science, language acquisition, natural language processing, and cognitive science -- will be presented. The issues dealt with by these networks will also turn out to be relevant to the discussion of CLASPnet in Chapter 3.