An Application of Probabilistic Grammars to Efficient Machne Translation

In this paper we present one of the algorithms used to parse probabilistic context-free grammars: the A* parsing algorithm, which is based on the A* graph search method. We show an example of application of the algorithm in an existing machine translation system. The existing CYK-based parser used in the Translatica system was modified by applying the A* parsing algorithm in order to examine the possibilities of improving its performance. This paper presents the results of applying the A* algorithm with different heuristic functions and their impact on the performance of the parser.