MODELIZATION OF AMBIGUITY RESOLUTION USING A CONNECTIONIST NETWORK

This paper presents a model of ambiguity (treated as a special case of polysemy) based on an explicit linguitic theory, using differential geometry as a mathematical framework and connectionist networks for computer implementation. This model can be used to undertake simulations that can be evaluated by psychological experimentations, specially concerning ocular movements during reading. The main claim is to bypass the opposition between the two main strategies in computer ambiguity processing (parallel or sequential with backtrack), corresponding to the two important psychological hypotheses about human ambiguity processing (single or multiple activation).