Last October, I started writing a new book that was supposed to be a collection of my research papers and doctoral dissertation. It is co-authored with my PhD advisor, Professor Zeng. The book titled ‘Social Multimedia Signals: A Signal Processing Approach to Social Network Phenomena’, is now available.
One of the biggest motivations for writing this book was to bring the signal processing/multimedia community and network science/sociology community together, and highlight topics where they could collaborate.
During the many drafts, we were made to strengthen, better comprehend and sometimes reassess certain aspects of how social networks behave — especially when you study them from a signal processing point of view (as opposed to pure structural features). Envisioning a network as a signal generating agent is initially counter-intuitive to most. Which is exactly why you should read this book!
From Networks to Signals
We have been very curious about the Social Web. There is hardly any debate that it has fundamentally transformed our digital lives. We witnessed a multitude of interesting online phenomena, including social media, the stream, click-baits, virality, memes, contagions, social-driven recommendations, social search and what not.
Social networks are enormous data generation engines. The essence of the Social Web is in how information spreads, via networks. Structure of the network appears to be a fundamental attribute of such systems. And this is how researchers have predominantly analyzed information dispersion in the Social Web — by its structure. Except, there is more to dispersion dynamics than just the structure.
By all means, network information diffusion respects the structure (you cannot defy it). But we need to study certain other parameters in order to completely understand the dynamics of information flow in social networks. Such parameters can be measured by monitoring multimedia behavior within the network, which manifests as signals.
The book is about all these “signals” — their occurrence, detection, estimation, origin, growth, decay, life-cycles and how their interaction leads to emergent phenomena in social networks.
The Utility of Transforms
So why should we study networks as signals, in addition to structure? Here’s a motivating example from the historic annals of science, where obscuring the fundamental attribute in data describing some system is profitable.
Consider the Fourier Transform. One reason it drastically changed the course of signal processing is rooted in the fact that it opened a door to analyzing time-series data by obscuring the fundamental dimension — time. In the Fourier domain, time is an ignored dimension.
Many ideas in this book strive to analyze networks similarly, by de-emphasizing (masking) the network’s most obvious and visible attribute — the structure. Instead we focus on the frequency of signals conveyed by multimedia behavior.
Multimedia - the lifeblood of social networks
Online social networks would be useless if we could not use it to communicate. And today, we communicate via multimedia. The term multimedia refers to multiple forms of media. And just like the streets of a city are filled with pedestrians, cars and buses, social networks have information channels that carry various types of multimedia — textual, audio, images, videos, gifs etc. from one node to another.
The way multimedia behaves in the Social Web can tell us a whole lot about the governing dynamics of information diffusion occurring within the networks. The behavior of multimedia is encoded in signals, and their detection/estimation enable us to quantify information dynamics of the network itself.
For example, here are the signals regarding the #Ferguson protests, based on how it trended in Twitter across several US cities. We can unravel critical dynamics of information networks and media diffusion in online societies by analyzing these signals.
These signals enable us to perceive information and attention dynamics in networks like never before; which is hard to achieved by pure structural analysis.
Unravelling the Ripple Web
It took us about 8 months to conceptualize and bring this book to ink. We had some strong material from our published research papers. But we also talked to several researchers in related communities, trying to gauge their research aspirations, frustrations and their vision of progress in their respective fields. We packed as much relevant information as possible about the current condition of the Social Web, the automated tools that mine it and research areas which will be extremely promising in future.
Our priorities were three key phenomena: (1) Social Networks, (2) Signal Processing, (3) Semantic Web.
If you are unsure how these phenomena are interrelated, the book could be valuable resource, since it presents each in the light of another and demonstrates the synergy between them. These three phenomenon together give rise to the Ripple Web, which is a core concept discussed in the book.
Comments/Thoughts/Questions : @_RoySD or email@example.com