Author profiling and Gender Identification have gained relevance in the last few years. The goal of the research in these fields is to extract certain demographic information on the authors of texts by analyzing their writing at several levels. In our work, we address the problem of the identification of the gender (male vs. female) of the authors of opinion pieces published online. Unlike the overwhelming majority of the proposals, we argue that the use of deeper linguistic features (i.e., syntactic and discourse structure), instead of mainly lexical features leads to a higher accuracy of gender identification. Using such features with supervised machine learning, we achieve very competitive results with accuracies over 84%.