The purpose of this study is to establish a method for identifying semantic relations in compound nouns in patent documents by using linguistic information. The information such as grammatical or semantic features plays a key role for analyzing semantic relations in compound nouns. We used the information about immediately succeeding case particles, adjective-forming suffixes and a verb suru "do" as grammatical features, concept classifications in EDR dictionary as semantic features. The performance of the fully automated statistical method was found to be good at 84% by using both grammatical and semantic features. This result shows advantage of our method. © 2008 Springer-Verlag Berlin Heidelberg.