The Study Of Extracting Feature Words On Chinese Search Engine

Along with the quick development of Internet, the information on Internet increases quickly everyday, howing to obtain the information that customer want from this enormous resources database,is an important problem that people face.The Search engine is the most valid means now. Express according to the latest inquisition of CNNIC in 2004 January, the Search engine shares 61.6% of the network service usage rate, the usage rate of the only next to E-mail. The importance of the search engine is very obvious.Although current Search engine turn to intelligence the direction develop, the most important index measured quality of Search engine——the precision and the recall, is still the popular study. At first,This paper studied the HOWNET.Generally, customer not only hope that he can get the information that contains the keywords, but also hope that get the information that relate to the keywords. According to this,the paper introduces how to compute the associativity between words adopting Dynamic "center of gravity forward lead". The system can make the comprehension to the keywords and association with the keywords making use of the associativity,that resolved the problem of matching keywords strictly. Finally, we propose the method of extracting feature words according to the associativity,and then adjust the weight of the feature words by BP artificial nerve network.