Applications of Social Media and Social Network Analysis
Edited by Kazienko, P. & Chawla, N.
Series: Lecture Notes in Social Networks
2015. Springer 240p
This edited collection includes papers on sentiment analysis, information diffusion, trend spotting, reputation metrics and network structures. Perhaps on first viewing some of these topics could seem to have limited relevance to IR, but in fact they deal with what are increasingly important constituent tools and techniques for IR over the social web.
Rather than considering the papers in order, I will proceed in the order of those I consider to be of most interest and relevance to the community.
The best studies for me had coupled ML/analytic techniques with the evaluation of interfaces from a UX perspective. I would say that the most interesting chapter doing this was from Smith et al [1] which presents a system (TopicFlow) for analysing and visualising emerging and evolving topics. This is done by LDA topic analysis on sets of texts divided into time slices with clustering between time slices to track similar topics over time. The interface was tested with users and largely validated as a means of trend-watching. A similar concern for “opening up the black box” motivated the paper from Michaelis et al [2] which looked to explain predictions made by a Bayes-net based system by allowing the user to drill into the features used for a prediction. I really liked this attempt to expose the system’s function and assess how well it worked for users.
Elsewhere, a need to consider time-sensitivity and varying network dynamics in the derivation and application of algorithms seemed to be a common theme. Lee & Oh [3] consider time-sensitivity and “velocity” in the derivation of reputation metrics in a social network. The approach seemed interesting, though the validation methods were not wholly convincing. In considering a physical transport network, Cheng et al [4] show that a dynamic centrality calculation improves on a static one and serves to highlight potential bottlenecks at different times of day. This might be a useful thing to take back into virtual networks.
Rather than adapting to temporal aspects, Albano et al show that it might be useful for the study of network diffusion to take normal time out of the equation altogether [5]. These authors propose “intrinsic” time which is based on network changes rather than external or “classic” time. Their case seems partly justified, particularly if comparing/contrasting different types of network. This paper was notable also for the use of some interesting large datasets (GitHub open source repository activity, Webfluence blogosphere networks, SocioPatterns human networking).
Also in a network diffusion vein, Bao et al [6] study the spread of rumour on social media and note that an “epidemic” model is insufficient for studying rumour as people either believe rumours or they don’t (ie. there are two types of infection). Their solution to nipping rumours in the bud are high-degree rumour squashing nodes. But I didn’t really see how this could work in practice short of forcing users to follow “big brother”-like rumour detection & squashing bots!
Two papers deal with sentiment analysis. Janssen et al look at building real-time sentiment analysis using cloud-based machine learning and relatively simple and broad-brush lexical and syntactic features of tweets [7]. By contrast, Lancieri and Leprêtre [8] focus more on building a custom sentiment lexicon for a domain based on the ground-truthing offered by positive and negative reviews. They include the intuition that negative reviews tend to be longer and more detailed!
Aside from the poor copy editing in places, I had two main criticisms of the collection overall. Firstly, several of the papers fail to indicate how the interesting approaches or effects described might realistically be implemented in practice. For me this undermines their credibility – if the researchers have not (even briefly) considered possible application and integration, this seems to be a commitment to obscurity. Secondly, with the quantitative and structural approaches taken by many of the papers, there seems to be a risk of a kind of “network determinism”. Are the social behaviours being studied a result of network constraints or are the structures themselves the emergent result of social behaviours? One or two of the authors seem to make the mistake of assuming the former and not considering the latter or the possibility of the mutual shaping of the two.
[1] A. Smith, S. Malik, and B. Shneiderman, “Visual Analysis of Topical Evolution in Unstructured Text: Design and Evaluation of TopicFlow,” in Applications of Social Media and Social Network Analysis SE – 9, P. Kazienko and N. Chawla, Eds. Springer International Publishing, 2015, pp. 159–175.
[2] J. Michaelis, D. McGuinness, C. Chang, J. Erickson, D. Hunter, and O. Babko-Malaya, “Explaining Scientific and Technical Emergence Forecasting,” in Applications of Social Media and Social Network Analysis SE – 10, P. Kazienko and N. Chawla, Eds. Springer International Publishing, 2015, pp. 177–192.
[3] J. Lee and J. Oh, “A Node-Centric Reputation Computation Algorithm on Online Social Networks,” in Applications of Social Media and Social Network Analysis SE – 1, P. Kazienko and N. Chawla, Eds. Springer International Publishing, 2015, pp. 1–22.
[4] Y.-Y. Cheng, R.-W. Lee, E.-P. Lim, and F. Zhu, “Measuring Centralities for Transportation Networks Beyond Structures,” in Applications of Social Media and Social Network Analysis SE – 2, P. Kazienko and N. Chawla, Eds. Springer International Publishing, 2015, pp. 23–39.
[5] A. Albano, J.-L. Guillaume, S. Heymann, and B. Grand, “Studying Graph Dynamics Through Intrinsic Time Based Diffusion Analysis,” in Applications of Social Media and Social Network Analysis SE – 6, P. Kazienko and N. Chawla, Eds. Springer International Publishing, 2015, pp. 103–124.
[6] Y. Bao, C. Yi, Y. Xue, and Y. Dong, “Precise Modeling Rumor Propagation and Control Strategy on Social Networks,” in Applications of Social Media and Social Network Analysis SE – 5, P. Kazienko and N. Chawla, Eds. Springer International Publishing, 2015, pp. 77–102.
[7] O. Janssens, R. de Walle, and S. Hoecke, “A Learning Based Approach for Real-Time Emotion Classification of Tweets,” in Applications of Social Media and Social Network Analysis SE – 7, P. Kazienko and N. Chawla, Eds. Springer International Publishing, 2015, pp. 125–142.
[8] L. Lancieri and E. Leprêtre, “A New Linguistic Approach to Assess the Opinion of Users in Social Network Environments,” in Applications of Social Media and Social Network Analysis SE – 8, P. Kazienko and N. Chawla, Eds. Springer International Publishing, 2015, pp. 143–158.