Information Retrieval Design Book Review

Information Retrieval Design. James D. Anderson and Jose Perez-Carballo. East
Brunswick, NJ: Ometeca Institute, Inc., 2005. 617 pp. $35.00. (0-9763547-0-5).
Information Retrieval Design is a textbook that aims to foster the intelligent user-centered
design of databases for Information Retrieval (IR). The book outlines a comprehensive
set of twenty factors, chosen based on prior research and the authors’ experiences, that
need to be considered during the design process. The authors provide designers with
information on those factors to help optimize decision-making. The book does not cover
user needs assessment, implementation of IR databases or retrieval systems, testing or
evaluation.
Most textbooks in IR do not offer a substantive walkthrough of the design factors
that need to be considered when developing IR databases. Instead they focus on issues
such as the implementation of data structures, the explanation of search algorithms and
the role of human-machine interaction in the search process. The book touches on all
three, but its focus is on designing databases that can be searched effectively, not the
tools to search them. This is an important distinction, despite its title this book does not
describe how to build retrieval systems. Professor Anderson utilizes his wealth of
experience in cataloguing and classification to bring a unique perspective on IR database
design that may be useful for novices, for developers seeking to make sense of the design
process and for students as a text to supplement classroom tuition.
The Foreword and Preface (by Jessica Milstead and James Anderson respectively)
are engaging and worthwhile reading. It is astounding that it has taken some 20 years for
anyone to continue the work of Milstead and write as extensively as Anderson does about
such an important issue as IR database design. The remainder of the book is divided into
two parts: Introduction and Background Issues and Design Decisions.
Part One is a reasonable introduction and includes a glossary of the terminology
that authors use in the book. It is very helpful to have these definitions early on, but the
subject descriptors in the right margin are distracting and do not serve their purpose as
access points to the text. The terminology is useful to have as the authors definitions of
concepts do not fit exactly with what is traditionally accepted in IR. For example, they
use the term “message” to refer to what would normally be called “document” or
“information object”, and do not do a good job at distinguishing between “messages” and
“documentary units”.
Part Two describes components and attributes of IR databases to help designers
make design choices. The book provides them with information about the potential
ramifications of their decisions and advocates a user-oriented approach to making them.
Chapters are arranged in a seemingly sensible order based around these factors and the
authors remind us of the importance of integrating them. The authors are skilled at
selecting the important factors in the development of seemingly complex entities, such as
IR databases. However, the integration of these factors, or the interaction between them,
is not handled as well as perhaps should be. Factors are presented in the order in which
authors feel they should be addressed but there is no chapter describing how the factors
interact. The authors miss an opportunity at the beginning of Part Two where they could
illustrate using a figure the interactions between the twenty factors they list in a way that
is not possible with the linear structure of the book.
What follows is a subject-based review. To aid the flow of the review and reduce
redundancy, chapters are grouped together by subject.
Chapters Two to Five: Scopes, Domains and Display Media (pp. 47 – 102).
Chapter Two describes the role and the value of the scope (i.e., the kinds of topical
questions an IR system must respond to) and the domain (i.e., the subject set in the
context of the work or life situation) within which IR databases must operate, and
differences between them. In Chapter Three the authors describe the type of documents
that users can retrieve using a database in terms of non-topical features (e.g., authorship,
media, intended audience). The authors suggest that document scope can be useful when
the search is constrained by document types (e.g., format, medium and periodicity), not
the topical content of the document. Chapter Four describes the document domain in
which the IR database must operate, the difference between primary and secondary
sources and how this document domain can be monitored and covered. In Chapter Five
the authors describe the media that can be used to present information to end-users. The
authors comprehensively list different forms of media that can be used in IR databases.
Chapters Six to Eight: Documents, Analysis and Indexing (pp. 103 – 176).
Chapter Six debates the issue of the optimal size of documents for retrieval from IR
databases. The authors stress the importance of selecting the appropriate document type
for retrieval. In Chapter Seven the role of the information indexed by the IR system is
described and a comprehensive overview of the types of information indexed is provided.
In Chapter Eight the authors describe how language text is indexed (either by humans or
automatically), although does not focus on the indexing of multimedia or other nonlanguage
texts. The authors spend much time on human indexing and supplement their
views with those of others. This is the first chapter in the book where traditional IR
techniques such as stemming, stopword removal and term frequencies are discussed.
Chapters Nine to Ten: Exhaustivity and Specificity (pp. 177 – 196). In Chapter
Nine the authors discuss “exhaustivity” (i.e., how well the topics and features of
documents are described). This is the first time traditional measures of IR effectiveness
such as recall and precision are mentioned. In Chapter Ten the authors outline the role of
“specificity” (i.e., the semantic term-document relation between the index terms and the
topics discussed in documents or the features of documents). The authors highlight the
ambiguity associated with the concept and describe other researchers’ views of it
Chapters Eleven to Thirteen: Displayed/Non-displayed Indexes, Syntax and
Vocabulary Management (pp. 197 – 364). Although not highlighted explicitly by the
authors, there is a marked shift between Chapters Ten and Eleven, and the role of the user
in the retrieval process become more obvious. In Chapter Eleven the authors examines
the role of displayed, browsable indexes for human inspection. This chapter leads
directly into the discussion of Syntax in Chapter Twelve and provides a reasonable (if
short) treatment of indexing display in IR and a summary of browsing in IR. Chapter
Twelve provides an extremely thorough discussion of the role of syntax as patterns for
putting terms together for indexing or searching. Chapter Thirteen describes the
management of vocabulary in IR, including the problems associated with differences in
vocabulary and the use of alphabetic indexes, thesauri and ontologies. The categorization
of searches offered in this chapter could be beneficial to readers, but perhaps the authors
could have given more attention to related work in this area.

Chapters Fourteen to Sixteen: Surrogation, Locators and Surrogate Displays (pp.
365 – 390). In Chapter Fourteen the content and uses of surrogates are defined (including
guidelines and standards for surrogates). Chapter Fifteen addresses locators, an essential
part of surrogates, and Chapter Sixteen describes the progressive presentation of
surrogate content from surrogate information to document full-text.
Chapters Seventeen to Eighteen: Arrangement and Size of Displayed Indexes (pp.
391 – 446). In Chapter Seventeen the book describes how indexes should be arranged for
presentation to users. Chapter Eighteen discusses the size of these displayed indexes. It
seemed unnecessary to have a separate chapter for each of these closely-related issues.
Chapters Nineteen to Twenty-One: Search Interface, Record Format and FullText
Display (pp. 447 – 536). The authors begin Chapter Nineteen by stressing that the
most important element in IR databases is the user interface. This is rather surprising
bearing in mind the amount of time the authors devote in the book up until this point
describing the role of components hidden from the user’s view. The review of research
in this area is sufficient for the purpose of the book, but I would have expected a more
detailed account of the role of Human-Computer Interaction in IR. Chapter Twenty
describes the record format that defines how the data will be tagged, labelled and stored
in the database. Chapter Twenty-One addresses the format of documents and how the
full-text should be presented to users.
Chapter Twenty-Two: Implementation and Evaluation (pp.537-541). The book
concludes with a brief discussion of issues in implementation, testing and evaluation.
Examples aside, this is the only time that these fundamental issues are addressed in any
way in the book. The chapter is a review of currently available commercial technologies
(e.g., EndNote, Reference Manager) rather than a description of how IR systems can be
developed. Since the book tries to make sense of the design space it may have been
useful to point readers to further reading that operationalizes many of the issues raised.
Testing and evaluation are covered briefly and organized by evaluation topic. The book
lacks a conclusion or takeaway message and would benefit from the inclusion of both.
The book uses examples to illustrate the concepts it describes. These include
examples of a book index, an indexing and abstracting service and new
encyclopaedia/digital library. These are used at the end of each chapter to reinforce the
concepts introduced therein and readers can follow the same three examples throughout
the book. These are useful for readers but also shows that the authors understand who
their readers are likely to be and the types of questions such readers would ask. The
“book index” example is somewhat trivial and is perhaps only useful to novices but the
others may be helpful to all readers. Surprisingly, aside from a few cursory references,
the book almost completely ignores the World Wide Web and its current role in
information dissemination. It may have been useful to include more references to this
domain, as many of the ideas discussed have been implemented there in some form.
The index of the book is comprehensive and the “related terms to consider” are
helpful and should have perhaps have been included for more than only a few terms. The
inclusion of section numbers after each citation in the bibliography is helpful in
pinpointing potentially useful parts of the book.
Those interested in using the book to design IR databases can work through the
chapters in the order provided and end up with a set of requirements for database design.
The steps outlined in this book can be rearranged in numerous orders depending on the
particular circumstances. This book would benefit from a discussion of what orders are
appropriate for different circumstances and how the requirements outlined interact.
I come away from Information Retrieval Design with mixed, although mainly
positive, feelings. Even though the aims of this book are made clear from the outset it
was still a disappointment to see issues such as implementation and evaluation covered in
only a cursory manner. The book is very well structured, well written and operates in a
part of the space that has been neglected for too long. The authors whet my appetite with
discussion of design and I would have liked to have heard a bit more about what happens
in requirements elicitation before the design issues have been identified and in
implementation after they have been addressed.
Overall, the book is a comprehensive review of previous research supplemented
by the authors’ views on IR design. This book focuses on breadth of coverage rather than
depth of coverage and is therefore potentially of more use to novices in the field. The
writing style is clear and the authors’ knowledge of the subject area is undoubted. I
would recommend this book to anyone who wants to learn about IR database design and
take advantage of the experience and insights of Anderson, one of the visionaries in the
field.