It becomes an increasing important task to discover desirable information from tremendous web resources

and therefore the more accurately the focused crawler describes the users’ interested topics

the more useful the information obtained by it will be.However

the neglect of context leads traditional approaches to the failure of achieving this goal.In order to overcome this shortcoming

a new context-sensitive topic description method based on ODP (Open Directory Project) is proposed in this paper.First of all

the best topic feature subset is determined by a new feature selection algorithm.And then

we optimize the topic description method to improve the performance by utilizing topic context.The experimental results show that this new approach extracts the features effectively with a better performance in both the precision and the sum of information while minimizing its dimension size significantly.