We propose a cross-language information retrieval method that is based on document feature modification and query translation using a dictionary extracted from comparable corpora. In this paper, we show the language-independent effectiveness of our document feature modification model for dealing with semantic ambiguity, and demonstrate the practicality of the proposed method for extracting multilingual keyword clusters from digital libraries. The results of our experiments with multilingual corpora indicate that our document feature modification model avoid the difficulties of language-/domain-dependent parameters.