3.1 Introduction

Biomedical terminologies and ontologies are frequently described as enabling resources in text mining systems ^{[1] [2] [3]} These resources are used to support tasks such as entity recognition (i.e., the identification of biomedical entities in text), and relation extraction (i.e., the identification of relationships among biomedical entities). Although a significant part of current text mining efforts focuses on the analysis of documents related to molecular biology, the use of lexical, terminological, and ontological resources is mentioned in research systems developed for the analysis of clinical narratives (e.g., MedSyndikate ^[4]), or the biological literature (e.g., BioRAT ^[5], GeneScene ^[6], EMPathIE ^[7], and PASTA ^[7]). Of note, some systems initially developed for extracting clinical information later have been adapted to extract relations among biological entities (e.g., MedLEE ^[8], GENIES ^[9], and SemRep/SemGen ^[10]). Commercial systems such as TeSSI, from Language & Computing (http://www.landcglobal.com/) also make use of such resources.

Entity recognition often draws on lists of entity names collected in lexicons, gazetteers, and, more generally, terminology resources. For example, lists of disease names can be easily extracted from disease resources, such as the International Classification of Diseases (ICD); from the disease component of general resources, such as the Medical Subject Headings (MeSH); and from specialized resources, such as the Online Multiple Congenital Anomaly/Mental Retardation (MCA/MR) Syndromes. In contrast, relation extraction may benefit from the relationships represented among terms in terminologies (e.g., Parkinson's disease child of Neurodegenerative diseases in MeSH), and in ontologies [e.g., Basal ganglia finding site of Parkinson's...