8.1 Introduction

Text mining is the extraction of information or knowledge from databases of literature, while data mining assumes databases of structured data as the source of knowledge. Contrary to the usual case in data mining, knowledge in text mining is assumed to be encoded in natural language text, which poses problems for computational analysis (see Chapter 2). It is desirable for text mining systems to be made aware of language, in order to identify effectively the structures of interest in the text.

Natural Language Processing (NLP) techniques have been successfully applied to fields that need to exploit natural language texts. Corpora, which are the collections of written (or spoken) material of a language upon which linguistic analysis is based, have always been at the heart of NLP research, providing the reference material to put flesh on the skeleton of theoretic models. Increasingly, annotations are made over corpora, to make explicit the structures of interest that implicitly reside in text, and to provide direct reference to these structures.

This chapter gives an overview of corpora and annotations that are available in the domain of biomedicine, and discusses related issues and problems.

8.2 Literature Databases in Biology

Literature databases contain a large amount of knowledge, and become knowledge sources to be mined for knowledge nuggets. This section introduces literature databases available in the biomedical research community. Copyright issues also are addressed, since text mining is inherently subject to copyright conditions.