Coding of natural language task descriptions prior to their classification by neural networks

Explores ways of converting natural-language documents from symbolic to subsymbolic form. This permits sentence processing to be performed by artificial neural networks at a more global and semantic level. Before processing natural language text with neural networks, it is necessary to extract the word parts which carry the meaning of each sentence. It is also necessary to standardize the form of each sentence before it is presented to a neural network. Processing of natural language with neural networks should improve the effectiveness of high-performance computers by making a switch from the traditional, sequential analysis to a parallel synthesis of the subsymbolic parts of natural language sentences