These days, recommender systems form an integral part of many information processing services on the Internet, especially on the web. In fact, recommendation systems help manage the continuous growth of information available online. Classical information search often involves a recommendation step, where search results are ranked or presented as recommended items according to the search query. However, recommendation is also used for domain-dependent tasks where visitors are guided through items that have the potential to attract either a particular visitor or whole groups of visitors. Recommendation is thus often connected with personalization--when items are recommended for particular users and/or their contexts specifically.

This book presents the results of research conducted in the course of a doctoral study on improving recommendations on the web. The study exploits the important concept of the social web, or user-contributed information. The author focuses on tagging data and proposes new methods that exploit user-provided tagging data to improve the accuracy of recommendations, as well as the quality of the corresponding explanations.

The author proposes the concept of user- and item-specific tag preferences, which makes this work suitable for specific domains where users are willing to provide details on their feelings (likes and dislikes) about particular items. Even though this can be a rather demanding assumption, there are more and more examples of sites where visitors provide just such detailed feedback (for example, about movies or reading resources).

The book contains seven chapters and a bibliography. The style of the book reflects its origins as a doctoral dissertation (presenting contributions to the state of the art, including an evaluation of each particular contribution). It seems that the author published the doctoral dissertation in its original form (including a list of publications in the introductory chapter, an explicit list of contributions, and an appendix discussing joint publications), so the book can also serve as a good example of how to write such a document. I like this work; without much effort, it could be converted to a monograph with more impact.

The first chapter presents the motivation and basic concept of rating items by attaching tags to them. Chapter 2 introduces the basics of recommender systems, including recommendation techniques and techniques for evaluating recommender systems. In 25 pages, the author provides a comprehensive and well-balanced survey. I recommend this chapter to any graduate student interested in recommender systems.

Chapters 3 and 4 describe the recommendation algorithms proposed by the author. The first is LocalRank, a tag-recommending algorithm that is scalable, in contrast to the state-of-the-art algorithm FolkRank. This algorithm focuses on just a small part of the user tags for resources while generating highly accurate tag recommendations. The second is a context-specific tag preferences algorithm, which is used to improve the predictive accuracy of recommendation systems.

Chapters 5 and 6 discuss the results in the explanation field, which directly influence the strength of the recommendation algorithms. Using the primary idea of this work, which is to exploit tags to improve a recommendation, the author analyzes several tag-based explanations. The analysis is conducted in two user studies that provide readers with methodological insight into this kind of research results evaluation.

The book concludes with chapter 7, which lists the key contributions of the doctoral dissertation and discusses limitations. A very short section presents future perspectives. In fact, I expected more discussion here.

I recommend this book to graduate students and researchers in the field of recommender systems and the social web. It can also serve as inspiration on how to conduct user studies for evaluating various information processing approaches.