Text this: Speaker verification method based on cross-domain attentive feature fusion