Science of Swarms

Finding Trends in Online Forums

Besides tracking when who is most influential, we can also track when those influencers talk about what products - when is the buzz about a certain product the greatest. Based on standard information retrieval procedures (tfidf) we can compute a concept map of the most significant terms used in the online forum, and the relationships between those terms. Figure 18 displays a concept map for the 50 most significant terms over the entire 4 months period.

Figure 19. Concept map (term view), most significant actors on term "bose" shown

In figure 19, the reader has selected the term "Bose", showing that "Randy Warren" is the most influential person talking about Bose.

Using the TeCFlow taxonomy feature allows us to identify the most significant messages about the brands we are most interested in. The taxonomy:
loudspeaker brand
Bose
Klipsch
Denon
Marantz
Polk
will categorize documents about the loudspeaker brands Bose, Klipsch, Denon, Marantz, and Polk into 5 clusters. Figure 20 displays the result of clustering the documents into these 5 categories.

Figure 20 Auto-categorization by taxonomy, and most significant document about "Polk"

Figure 20 also show a pop-up window brought up by the user, which lists all the documents about "Polk", sorted by significance. The most highly-ranked document tells that Chris "would take BA over Polk any day", an open invitation for Polk to get back to Chris and ask him why he got this unfavorable opinion about their product.

Finally, figure 21 shows us when in time what brand was the hottest. For example, discussion about paradigm speakers peaked around day 40. This means, at this point in time group betweenness centrality was high (the light yellow line), as was betweenness centrality of term "paradigm". Discussion about Denon, on the other hand, was relatively flat for the first ninety days, but became the dominant issue in days 92 to 120.

Figure 21 Evolution of Betweenness over time of key brands, identifying trends

Latest from the Blog

Predicting the World Cup 2010 Winner — Recently I was giving a presentation about Coolhunting through Swarm Creativity at the CRM Forum in Zurich. As there was the Soccer World Championship at that time in South Africa, at the...… Read More

Predicting Stock Market Indicators Through Twitter "I hope it is not as bad as I fear" — We have been working on trying to predict market indicators for quite some time by analyzing Web Buzz, predicting who will win an Oscar, or how well movies do at the box...… Read More

Read More from the Blog

Subscribe to the RSS Feed
  • The MIT association is very reassuring, their approach was very methodical, confident, patient, helpful and encouraging. Outstanding team, focused, conceptually easy, quick at negotiating time and delays, great class, simple and user friendly despite the complex systems.
    Maggie Palazzolo, Research Coordinator
  • Excellent experience overall. Correspondence was kind, fast and helpful mostly via email; we suggested new features as needed and they were added quickly.
    Daniel Oster, Assistant to the CEO
  • There are many others who say they can do similar things, but I don't think they are actually capable. GalaxyAdvisors is different because many others don't use the social network methodology to learn what people are talking about.
    Daniel Oster, Assistant to the CEO
  • Very Unique. There are no others like them! We trust galaxyadvisors and feel that relationship is critical, as we are not social network analysis experts and need their help and understanding.
    Mirjam Hauser, GDI Analyst