Clinical Information Retrieval with Split-layer Language Models

the increasing prevalence of electronic medical records (EMRs), search technologies for these systems hold signifi- cant promise for improving patient and population care. We present a split-layer language model that embeds linguistic layers from existing NLP systems in retrieving medical docu- ments. On the cohort identification task of the TREC Med- ical Records Track, our approach shows improvement over baselines, with the best performance achieved by mixing in all tested layers of NLP artifacts.