Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Review Help
Search
Prediction and inference from social networks and social media
Kawash J., Agarwal N., Özyer T., Springer International Publishing, New York, NY, 2017. 225 pp. Type: Book (978-3-319510-48-4)
Date Reviewed: Nov 30 2017

Traversing the social graph of social media networks has become the holy grail for basic research into the human condition. This book contains in-depth discussions of how social networks (SNs) enable interaction between participants and insights into “medicine, business, education, politics, and activism” (p. v). Those insights are gleaned from algorithms dealing with the massive data accumulation in online social media (OSM). Written by different authors, the book is a multifaceted discussion of SN topologies and methodologies describing different ways to harness the data from OSM to make predictions and inferences.

Chapter 1 outlines a feature matrix for assessing emotions in OSM along the lines of positive, negative, or neutral to predict the mood of users. Using a large corpus of Twitter feeds, the authors extract features such as engagement, gender, and common linguistic markers to achieve individual mood predictions based on temporal information.

Chapter 2 details the constructions of a social network for indexing medical images in a multilingual environment. This chapter includes a very thorough discussion of constructing an SN from the ground up. Experiments support the veracity of their claims of how to successfully integrate multilingual diagnoses in one medical SN.

Chapter 3 investigates the status of members belonging to different yet overlapping networks. The authors present an overlapping community detection (OCD) algorithm based on two social dynamics: “signed disassortative degree mixing and information diffusion for signed networks” (p. 52).

Chapter 4 provides a framework for using link prediction as a possible indicator for future illnesses based on correlating symptoms and age. After establishing a “weighed symptom network” consisting of values (for example, high, low, abnormal, or normal) for medical parameters (for example, blood count or urinalysis), the authors use unsupervised link prediction to presage age-related health conditions.

Chapter 5 looks at link prediction for recommendation systems such as e-commerce sites suggesting consumer goods based on previous purchases. It discusses algorithms that find hidden links between users and items. For efficiency, a graph database model (as opposed to a relational model) is used.

Chapter 6 looks at quality control for Wikipedia contributions. Ultimately, the authority of the reviewer, the amount of his or her contribution, and the relation between the author and the reviewer play a significant role in establishing the degree of veracity of actual Wikipedia articles.

Chapter 7 investigates Twitter activity before the London riots in conjunction with weak signals that surfaced on OSM well before the event happened. Identifying those weak signals as part of collective behavior could predict large-scale phenomena. The authors use established techniques such as keyword, frequency, and sentiment analysis, and add in this chapter a machine learning model for tweet classification.

Chapter 8 discusses a knowledge exchange between members of massive open online courses (MOOCs). Since the large number of participants in a MOOC cannot be served individually by a tutor, discussion forums are the best way to exchange information between participants. The sublanguage character of the forums allows for a more focused approach to investigate knowledge exchange, which can be improved by better Q&A channels and directed pairing of information seekers with givers.

Chapter 9 addresses the dichotomy of unique versus multidimensional SNs (MSNs). MSN modeling assumes that the different roles someone plays in society can be represented in an SN as well. A participant can be a neighbor in one dimension and a friend in another dimension, and the behavior can be markedly different in each role when it comes to information propagation.

The book is geared toward a technical audience with an assumed appreciation of statistics and mathematics. Monetization of SN is not a primary issue in this book. Chapter 7 stood out because of the unexpected predictability of social events long before they occur based on tweet classifications. Chapter 1 showed a surprising correlation between mood prediction and temporal information using an impressive dataset to back up the analysis. Chapter 5 showed the practicality of a graph-based database model. Chapter 4 seemed to have the most immediate pay-off for an important issue in our society: healthy aging.

All nine chapters of the book deal with prediction and inference from totally different vantage points in a very compact manner. Hence, an index to help the reader cross-reference topics would have been helpful. The number of typographical errors and errant word choices is noticeable (for example, Turkley/Turkey (p. vii), activates/activities (p. 1), emotions/emoticons (p. 31), diagonisis/diagnosis (p. 40), and wealthy/rich (p. 46).

The book, part of the “Lecture Notes in Social Media” series, addresses a panoply of topics, but also coalesces around methods such as link predictions, information propagation, and MSNs applied to health, learning, and marketing. It was surprising that in a book on prediction none of the authors mentioned the “friendship paradox” and its ramifications for propagation in networks. The friendship paradox [1] formulated by Scott L. Feld in 1991 originally stated that on average most people have fewer friends than their friends have. In 2009, Christakis and Fowler [2] applied the friendship paradox to successfully predict influenza outbreaks at Harvard. Multidisciplinary research is required to successfully analyze the massive data aggregation in OSM to improve the human condition.

Reviewer:  Klaus K. Obermeier Review #: CR145686 (1802-0051)
1) Feld, S. L. Why your friends have more friends than you do. American Journal of Sociology 96, 6(1991), 1464–1477.
2) Christakis, N. A.; Fowler, J. H. Social network sensors for early detection of contagious outbreaks. PLoS ONE 5, 9(2010), e12948.
Bookmark and Share
  Reviewer Selected
Featured Reviewer
 
 
Web-Based Interaction (H.5.3 ... )
 
 
World Wide Web (WWW) (H.3.4 ... )
 
 
Multimedia Information Systems (H.5.1 )
 
 
Systems And Software (H.3.4 )
 
Would you recommend this review?
yes
no
Other reviews under "Web-Based Interaction": Date
Building knowledge communities with webs of connections
Smith B. In Knowledge management. Norwood, MA: Artech House, Inc., 2002. Type: Book Chapter
Dec 18 2002
Collaboration via internet and Web
Tomek I. In Computer science in perspective. New York, NY: Springer-Verlag New York, Inc., 2003. Type: Book Chapter
Apr 6 2004
XForms: XML powered web forms
Raman T., Addison-Wesley Longman Publishing Co, Inc., Boston, MA, 2003.  304, Type: Book (9780321154996)
Mar 23 2004
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