Question answering (QA) is a type of information retrieval in which a computerized system performs complex intelligent processing in response to queries posed in natural language (or a reasonable variation thereof). The complex intelligent processing involved in QA (compared to other forms of information retrieval) is typically characterized by some form of semantic analysis of the query, a search strategy that considers interrelationships between concepts in queries and in the corpus being searched; selection and prioritization of the contents of the response according to the query; and formulation of the response in terms appropriate to the query. Question answering has long been an area of research in artificial intelligence and information retrieval, and has gained increasing importance with the advent of the World Wide Web.

This book grew out of the US intelligence community’s advanced research and development activity in information technology (ARDA) advanced question answering for intelligence (AQUAINT) program. The volume is largely a compendium of research reports on specific projects or problems, prepared by experienced researchers in the field.

The bulk of the book is taken up by 18 chapters on different aspects of QA, collectively covering a wide variety of activity in the field, ranging from question answering for terminology and definitions to advanced areas such as summarization, inferencing, and reuse. An introductory chapter provides an overview and roadmap, and two further chapters contain a short discussion and survey of past and present approaches, and of question answering on the World Wide Web, respectively. The rest of the book is divided into six sections, each a compilation of research reports on one facet or another of recent research in the field. The individual contributions are of high quality and reasonably deep, in spite of the limited amount of space for each; the papers (generally 10 to 14 pages apiece) are closer to conference length than journal length, most originating in an American Association for Artificial Intelligence (AAAI) symposium.

Being largely a collection of selected research papers, this book is a fairly comprehensive (but not necessarily complete) survey of the field. Readers desiring further information will need to consult other sources as well (for example, the text retrieval conference (TREC) series). The book is, however, an excellent starting point, and well worth the reader’s effort. The intended audience (researchers, practitioners, and students) will find it helpful, and it should also be suitable for a university seminar course. Prior exposure to artificial intelligence, information management, information retrieval or related subjects would be particularly helpful if the reader is to get the most out of this book.

A comprehensive glossary and sizeable bibliography are included. Production quality is generally satisfactory, though deficient in places, especially in some of the figures. For example, the roadmap on page 13 is rather difficult to comprehend, due to the reduced size of its contents, as are the contents of Table 7.1 (page 90), and the graphs in Figure 14.2 (page 188). Overall, this book will make a good addition to an information management/retrieval or artificial intelligence collection.