USING SINGLE-LAYER NETWORKS FOR DISCRETE, SEQUENTIAL DATA - AN EXAMPLE FROM NATURAL-LANGUAGE PROCESSING

Natural Language Processing (NLP) is concerned with processing ordinar y, unrestricted text. This work takes a new approach to a traditional NLP task, using neural computing methods. A parser which has been unsu ccessfully implemented is described. It is a hybrid system, in which n eural processors operate within a rule based framework. The neural pro cessing components belong to the class of Generalized Single Layer Net works (GSLN). In general, supervised, feed-forward networks need more than one layer to process data. However, in some cases data can be pre -processed with a non-linear transformation, and then presented in a l inearly separable form for subsequent processing by a single layer net . Such networks offer advantages of functional transparency and operat ional speed. For our parser, the initial stage of processing maps ling uistic data onto a higher order representation, which can then be anal ysed by a single layer network. This transformation is supported by in formation theoretic analysis. Three different algorithms for the neura l component were investigated. Single layer nets can be trained by fin ding weight adjustments based on (a)factors proportional to the input, as in the Perceptron, (b)factors proportional to the existing weights , and (c)an error minimization method. In our experiments generalizati on ability varies little; method (b) is used for a prototype parser. T his is available via telnet.