Genetic programming for natural language parsing

The aim of this paper is to prove the effectiveness of the genetic programming approach in automatic parsing of sentences of real texts. Classical parsing methods are based on complete search techniques to find the different interpretations of a sentence. However, the size of the search space increases exponentially with the length of the sentence or text to be parsed and the size of the grammar, so that exhaustive search methods can fail to reach a solution in a reasonable time. This paper presents the implementation of a probabilistic bottom-up parser based on genetic programming which works with a population of partial parses, i.e. parses of sentence segments. The quality of the individuals is computed as a measure of its probability, which is obtained from the probability of the grammar rules and lexical tags involved in the parse. In the approach adopted herein, the size of the trees generated is limited by the length of the sentence. In this way, the size of the search space, determined by the size of the sentence to parse, the number of valid lexical tags for each words and specially by the size of the grammar, is also limited.