General descriptions in medical records are so diverse that they are usually entered as free text into an electronic medical record, and the resulting data analysis is often difficult. We developed and implemented a template-based data entry module and data analyzing system for general descriptions. We developed a template with tree structure, whose content master and entered patient's data are simultaneously expressed by XML. The entered structured data is converted to narrative form for easy reading. This module was implemented in the EMR system, and is used in 35 hospitals as of October, 2006. So far, 3725 templates (3242 concepts) have been produced. The data in XML and narrative text data are stored in the EMR database. The XML data are retrieved, and then patient's data are extracted, to be stored in the data ware-house (DWH). We developed a search assisting system that enables users to find objective data from the DWH without requiring complicated SQL. By using this method, general descriptions in medical records can be structured and made available for clinical research.