This study focuses on prospects of using data-driven learning (DDL) methodology in teaching German for specific purposes (GSP) at Peter the Great Petersburg Polytechnic University. Unlike General English and English for specific purposes that are successfully taught with DDL worldwide, DDL does not seems to have paved the way to the GSP classroom yet. The paper analyses the reasons for this. One of the obstacles is that large general corpora do not contain field specific terminology, whereas specialized German corpora are not available. German scientific termhood is featured with high portion of compounds. Preliminary study of a small corpus of scientific papers built by the authors for this research proved this fact which makes it relevant to academic needs of learners of German who make a lot of mistakes in using and translating German compounds. The article presents lists of the most frequent nouns and some of the most frequent compound nouns that contain the nouns from the first list as stems generated using the AntConc freeware corpus analysis toolkit and a set of corpus-based teaching materials to help students to handle this lexical category. Further research should confirm the efficiency of DDL in teaching German compounds in the GSP classroom and effectiveness of the described way of integrating DDL in the GSP classroom: determining problem areas (gaps) in students’ knowledge -> finding these structures in the corpus -> development of corpus-based teaching materials, and using them in the GSP course.