Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Review Help
Search
Modeling intention in email : speech acts, information leaks and recommendation models
Carvalho V., Springer Publishing Company, Incorporated, Berlin, Germany, 2011. 116 pp. Type: Book (978-3-642199-55-4)
Date Reviewed: Jun 18 2012

This monograph describes the author’s research on problems related to email management. Specifically, the book presents various machine learning techniques that can be used to address different issues of email exchange in a professional environment. Existing email messages are considered as potential data that can train efficient tools for prioritizing messages, protecting against information leaks, coordinating shared tasks, and tracking commitments and deadlines.

The author presents a variety of machine learning methodologies and, most importantly, considers their combination in a very complex research field. Indeed, this research area has two major problems: the availability of data, due to the personal and confidential nature of the email messages, and the complicated nature of data, since the raw material is stored as natural language texts and the messages are linked in various ways. It is important to note that the research described in the book concerns open problems that have not been addressed adequately up to now, so the prospect of future research in this area is very good.

The targeted audience consists mainly of computer scientists, either researchers in machine learning or professionals who use email services in their work. The book would also be useful in education, especially to generate assignments for advanced students. Readers should have a machine learning background, but in general this monograph is easily readable. It provides a brief but comprehensive introduction to the peculiarities of email problems and the appropriate methods for addressing them. The updated bibliography and related work sections are useful for further study.

The material is organized in 6 chapters and two appendices. Each of the main chapters contains methods and results related to a specific problem, so each chapter can be studied independently.

The first chapter introduces the main topics of the book, briefly discusses the motivation of the research, and highlights the importance of email in everyday activities.

Chapter 2 deals with email speech acts, noun-verb pairs expressing intention (for example, request, commitment, or proposal). A taxonomy scheme of such acts is proposed to facilitate classification algorithms that efficiently categorize emails. This categorization is obviously a very important task for the automatic management of emails. The taxonomy and various classifiers are applied to an email corpus, and the results are evaluated and compared through extensive experiments. The treatment of the sequential correlation of emails, the collective classification algorithm that takes into account the links between them, and the linguistic analysis that examines sequences of words to improve classification are especially interesting.

Chapter 3 deals with another crucial aspect of email communication: information leaks due to misdirected messages. The author’s approach to the problem is based on outlier detection, aiming to prevent the sending of an email to an unintended recipient by detecting significant differences from past messages sent to that recipient. The difficult problem of finding real cases of unintentional transmission is discussed here and is addressed by simulating synthetic leaks in a large email dataset (the Enron dataset). The proposed methods produce rankings of recipients. Techniques like cosine similarity and k-nearest neighbors utilize the textual contents of emails. Furthermore, social network information, like numbers of received and sent messages and concurrences of recipients, are used to improve the results. There is also an interesting attempt to discover and study real cases of email leaks.

Chapter 4 deals with another important problem: the use of intelligent algorithms for recipient recommendation. This a very useful potential facility in professional email exchanges since it can recommend (to the author of an email) possible recipients based on content and past recipients. The Enron email dataset is used to train and evaluate the algorithms. Probabilistic expert search models are used to predict candidate recipients, while the recommendation is addressed as a multi-class classification problem. Other issues, such as email address auto-completion, are also discussed in this chapter.

In chapter 5, the author describes a user study implementing recipient recommendation and leak detection algorithms in an email client. Specifically, Cut Once, a Mozilla Thunderbird extension, was developed to be used and evaluated by humans in a realistic environment. The human subjects used the tool and provided feedback via questionnaires, which were subsequently analyzed. The results showed very positive results and provided valuable hints for future improvements.

A brief chapter 6 summarizes the previous chapters. The book also contains two appendices. The first contains guidelines of how to label email acts with specific verbs and nouns, while the second contains supporting material for the Cut Once user study.

In conclusion, this monograph provides a useful account of intelligent methods for addressing interesting and critical practical problems in email management. It is recommended mainly for researchers and practitioners in the field, but is also for those in fields involving human interaction through written text.

Reviewer:  Lefteris Angelis Review #: CR140277 (1210-1010)
Bookmark and Share
 
Electronic Mail (H.4.3 ... )
 
 
Classifier Design And Evaluation (I.5.2 ... )
 
 
Human Factors (H.1.2 ... )
 
 
Similarity Measures (I.5.3 ... )
 
 
Text Processing (I.5.4 ... )
 
 
Model Development (I.6.5 )
 
Would you recommend this review?
yes
no
Other reviews under "Electronic Mail": Date
Multimedia electronic mail
Borenstein N. Communications of the ACM 34(4): 117-119, 1991. Type: Article
Aug 1 1991
The Z-Mail handbook
Nelson H., O’Reilly & Associates, Inc., Sebastopol, CA, 1991. Type: Book (9780937175767)
Oct 1 1992
Electronic mail
Panko R., John Wiley & Sons, Inc., New York, NY, 1984. Type: Book (9780471262336)
Jun 1 1985
more...

E-Mail This Printer-Friendly
Send Your Comments
Contact Us
Reproduction in whole or in part without permission is prohibited.   Copyright 1999-2024 ThinkLoud®
Terms of Use
| Privacy Policy