In the field of biological science, correct field identification of species is of critical importance and requires effective access to and use of specialized biological information. Field guides, floras and faunas, and other publications are written to provide critical information for identification. However, full-text retrieval systems are ineffective at providing access to the very specific information needed for species identification. In this study, dichotomous key-like information that is often used in species identification is automatically generated from botanical documents and used to enhance full-text botanical document retrieval. Such information is used to describe document content and is also used to provide a form-based search that allows users to apply their domain knowledge to define various plant characteristics in their queries. Information extraction techniques were used to generate the key-like information from semi-structured botanical documents. Experimental results indicate that this approach improved search performance as well as user's satisfaction.