关键词: 短语识别，遗传算法，隐马尔可夫模型，Viterbi算法，层次分析

Abstract: To increase further the accuracy of noun phrase(NP) identification and utilize features of the genetic algorithm(GA) and the hidden markov modcl(HMM)，a novel HMM identification method based on GA was proposed. The method was based on a high-performance POS(parts of speech) tagging. During the training phase, model parameters were gained by the genetic algorithm. And during the identifying phase, an improved Viterbi algorithm for dynamic programming was first presented to identify the same hierarchy noun phrase, then the combination method of hierarchical syntax parsing and Viterbi algorithm was brought forward to identify those recursive noun phrases. Experimental resups show that this combined algorithm has achieved a high precision and recall rate of 94. 78 0 0 and 94. 29 0 0，rcspcctively,fully inosculating the strength of genetic algorithms and hidden markov model. This proves that the combination method has much better identification effect than the unitary hidden markov model identification approach.

Key words: Phrase recognition, Genetic algorithm, Hidden markov model, Viterbi algorithm, Hierarchical analysis