This post is first in a series that looks at the inferences you can make using social media data. The current analysis is a by-product of the Me Meme Twitter client that recently soft launched. Me Meme organizes your timeline into separate feeds based on your interests, helping you conquer information overload. It leverages Panoptez, a hosted Model as a Service from Zato Novo.
Prevailing wisdom says that who you associate with says a lot about you. By analyzing your social media graph, I can confirm that this claim has merit. Remarkably, it can also reveal non-obvious interests that provide insights into a person. This knowledge can be used for marketing, sales, or even plain-old relationship building.
First a bit of background. Panoptez, is a suite of models that I’ve developed and am licensing. They are components that are used to build custom data-driven applications. (See The Models as a Service Era has arrived for more information on this idea.) An example is the interplay between Panoptez and Me Meme. Panoptez provides the intelligence of extracting groups of people with similar interests and generating representative keywords that describe those groups. Me Meme uses this model output to display a Twitter timeline in separate feeds, so that each feed only contains tweets by the people in the group. Like a newspaper organized by sections, your feed becomes organized based on your individual interests. So Panoptez makes inferences, and Me Meme applies it to a specific problem.
What’s fascinating is the level of insight that can be gleaned from something as simple as a social graph. Take, for example, the NYC venture capitalist Matt Turck. What does his social graph say about him? Here are some of the groups Panoptez found within his social graph.
|Group labels||Number of members||Sample user – description|
|founder ceo, new york, venture capital||639||slazar – Doing what I can to help technology start-ups break through the noise as a Principal at General Catalyst (@gcvp) in NYC.|
|big data, design data, founder ceo, founder cto||115||Kellblog – CEO of Host Analytics, the leader in cloud-based enterprise performance management. See FAQ on Kellblog for disclaimers, including list of affiliations.|
|big data, data scientist, data science, chief data scientist, data analytics||114||statpumpkin – Director of Statistics Research – AT&T|
|3d printing, raspberry pi, art hacking, industrial revolution||29||MarleenVogelaar – Co-Founder Shapeways | 3D Printing | Innovator | Track and Field|
|bloomberg lp, bloomberg link, social media||26||evghood – Austin born. TCU grad. Current New Yorker. Works at Bloomberg LINK @bbglink. Small dog named Bo. Married to @cwhood.|
|internet of things, smart home, smartphone apps||22||iotwatch – founder @GNLteam / consultant at @designswarm / editor @ConnectedIOT / CEO @tinkerlondon /organiser @iotlondon @liquid_con @iotangels @techcityiwd / FRSA|
|founder ceo, frances startup, frances startup blog||21||LaFrenchTech La #Frenchtech est le nom collectif pour désigner tous les acteurs de l’« écosystème de #startups » français. #innovation #france|
These groups appear to isolate Matt’s interests. But how do we know for sure that these interests are accurate? Since I don’t know Matt personally, I can’t just ask him. Thankfully Matt blogs and has an informative bio that we can use a form of validation:
So his bio validates a number of the groups identified by Panoptez. This is further confirmed by his blog posts:
- The French Startup Ecosystem: At a Tipping Point
- The State Of Big Data in 2014: a Chart
- Introduction to the Internet of Things (Video)
- The Rise of the Female Hardware Entrepreneur
What have we learned from this post? For a given user, we’ve managed to extract key interests of that user by simply analyzing the user’s social graph. Throw in a dash of natural language processing to generate summary labels for the groups, and we have a powerful model for understanding users. All of this can be provided as a simple hosted model/data service that requires no infrastructure build-out nor a data science team. Contact me at firstname.lastname@example.org to learn more about Panoptez and Me Meme.
For those interested in leveraging Panoptez, there are two ways to access it (after subscribing to the service). The preferred approach is via a REST API and data feed. Alternatively, it is possible to work with the Panoptez R library directly. This analysis only requires three function calls:
> uc <- user_community(user.id) > list_communities(uc) 0 1 2 3 4 5 6 7 8 9 10 8 21 4 9 639 26 29 22 115 114 6 > head(print_community(1,uc),10) location name screen_name description 75334187 Paris Tatiana jama Tatianajama Co Founder & CEO @Selectionnist @Dealissime (acquired by LivingSocial) // Partner @50Partners @PartechEntrepreneur #BA #Passionate about startups 560454290 San Francisco, Paris Algolia algolia Algolia is a powerful Search as a Service API built for developers to deliver relevant results in their apps and website. 41675415 Paris, London, Europe Roxanne Varza roxannevarza startup lead @microsoft / @msftventures. cofounder @tech_eu @girlsintech_uk @gitparis & @failcon paris. ex techcrunch france editor @TCFR. epilepsy advocate. 4354801 Paris Marie Ekeland bibicheri VC @Elaia_Partners, Co-President @FRDigitale 5870292 NYC, Paris Frédéric Montagnon fred_montagnon Founder & CEO at Secret Media Inc. Entrepreneur. French. Engineer. Ex Codanova, OverBlog, Nomao, Ebuzzing. Blues, monoski and herbal tea addict. 116907977 New York French Morning FrenchMorningNY Le twitter du 1er web magazine des Français d'Amérique. 38021678 Paris – Montreuil Fleur Pellerin fleurpellerin Secrétaire d'Etat au commerce extérieur, à la promotion du tourisme et aux Français de l'étranger 2205307848 France La French Tech LaFrenchTech La #Frenchtech est le nom collectif pour désigner tous les acteurs de l’« écosystème de #startups » français. #innovation #france 267794397 Paris, France Liam Boogar LiamBoogar Editor @RudeBaguette - France's Startup Blog - We cover Web, Mobile & Tech in the Paris startup scene, balancing breaking news with in-depth analysis 112671889 Everywhere Kima Ventures kimaventures Xavier Niel @xavier75 and Jeremie Berrebi @jberrebi Angel Seed Fund