In research on question classification in question answering system

combining features can greatly help construct efficient question classifier.In order to deal with the problem of low performance of existing methods

a new method of diversity and importance based feature combination(DIFC) is proposed.By calculating the diversity between candidate feature and current combination for error and correct classification respectively

and the importance of candidate feature

features can be dynamically selected from candidate feature set.The experimental results of bag-of-words binding features on the HIT Chinese question set show that

compared with other methods

the new method is flexible and efficient

and gets more optimal feature combination.