With the MobileWorks crowdsourcing platform, anyone can analyze large sets of data that are simple for humans, but difficult for computers to understand. In the spirit of the upcoming U.S. 2012 Presidential Election, we initially ran a small experiment to show how – without any programming knowledge – anyone could build a high quality, data driven, analysis on a subject. For this experiment we chose to show America’s play-by-play reactions to this year’s presidential election.
For this initial test we collected 900 tweets (data points) between September 10th and September 11th, 2012. We sent this data via our BrainPower API to our distributed workforce to answer two questions per tweet for us. First question: “What political party is the tweet about?” Of the 900 tweets, 39% were categorized as either about the Republican Party or Democratic Party. The rest were about other parties or other aspects of the election not related to a specific party or not about the election at all. Second question: “What is the sentiment of the tweet?” Is the tweet Positive, Negative, or Neutral about the Political Party? The graph below best shows the results.
- Rather than a catch-all reply to any @mention, an intelligent reply could be generated within a specific set of options
- Use this same process to analyze the sentiment of any subject, including your company’s twitter followers
- Marketing to potential customers with intelligently designed messages: “Does anyone know when Metallica will be in SF again?”, your event ticketing company twitter account replies with “Hi @MetallicaFan Buy your Metallica tickets for SF – February 1st, 2013 at TheBestTicketingPlace.com/Metallica-SF/”
We will continue to gather and analyze data up until the day after the 2012 U.S. Presidential Election and periodically update our blog with analysis. If you are interested in receiving updates subscribe to our mailing list. Comment below with any questions, ideas for other twitter analysis, or if you just think its cool!
– Florian Mettetal