Medical text can present unique challenges for natural language processing (NLP) including documents that are several pages long, linguistic dependencies over different report sections, and only a tiny fraction of the report text being relevant to a specific information retrieval task. Many out-of-the-box deep learning architectures adapted from general NLP tasks are unable to efficiently tackle these challenges. This talk will cover the foundations of deep learning for NLP and then discuss how these architectures can be adapted for long medical texts. This talk will focus on the application of cancer pathology reports, but the techniques are applicable towards general biomedical and scientific NLP tasks in which specific information must be extracted from long documents.