The process of syntactic pattern recognition consists of two main phases. In the first one the symbolic representation of a pattern is created (so called primitives are identified). In the second phase the representation is analyzed by a formal automaton on the base of a previously defined formal grammar (i.e. syntax analysis / parsing is performed). One of the main problems of syntactic pattern recognition is the analysis of distorted (fuzzy) patterns. If a pattern is distorted and the results of the first phase are wrong, then the second phase usually will not bring satisfactory results either. In this paper we present a model that could allow to solve the problem by involving an uncertainty factor (fuzziness/distortion) into the whole process of syntactic pattern recognition. The model is a hybrid one (based on artificial neural networks and GDPLL(k)-based automata) and it covers both phases of the recognition process (primitives’ identification and syntax analysis). We discuss the application area of this model, as well as the goals of further research.