Chemical information mining: facilitating literature-based discovery

Debra Banville (ed) 

Boca Raton, US: CRC Press 2008 | 208pp | ?63.99 (HB) 

ISBN 9781420076493 

Reviewed by Peter Willett 

The effectiveness of systems such as SciFinder for searching the chemical literature arises from the detailed and costly human indexing that is carried out on the texts of journal articles, patents etc before they are put into a database. 

This book discusses the use of text-mining techniques that would enable such operations to be carried out by purely computational means. The techniques are clearly attractive in terms of cost-effectiveness but would appear, from this volume, to require considerably more research before they can provide a realistic alternative to human processing. Two chapters - Automated identification and conversion of chemical names to structure-searchable information  and Identification of chemical images and conversion to structure-searchable information  - stand out, providing critical, wide-ranging reviews of their respective topics that I am sure that I shall return to in the future. 

The remaining chapters are more varied in character. For example, the first two could usefully have been conflated into one, since they serve mainly to introduce the remainder of the volume. One of the chapters is more about XML in general than it is about chemical applications. Another on the Semantic Web, although clearly authoritative in nature, contains acronyms and computational terminology that hinder ready comprehension by the non-specialist. 

This book would have benefited from stronger editorial control to bring the various elements together, since there is some overlap of material, four different reference formats, and a very poor index. Nevertheless, it does provide an introduction to technologies that are likely to enhance our ability to access chemical literature in the future.