ABSTRACT
Machine Translation of natural languages (MT) has been tackled in many ways. This paper 1 discusses the usefulness of feed-forward neural networks to perform such tasks. Most conventional approaches to MT fail adequately to address the problem of allowing an MT system to learn from a human expert translator during its use. Most systems have hard-coded rules (with varying depth of meaning) that can be modified only by a knowledge engineer who is an expert in expressing such rules in the language of the computer. Neural networks, having the ability to learn from examples, are a possible solution to this problem. Research has already shown the usefulness of neural networks in various natural language processing tasks: Allen (1987), Jain (1991), Waibel (1988) and Waibel et al. (1991). If such a neural-network translation system wrongly translates a sentence it can be corrected and taught the proper translation by a user without any expert knowledge of how the computer stores and represents rules. This paper demonstrates the utility of neural networks in precisely this area on a small-scale translation problem.
