Our Intelligence department has developed a subject-detection algorithm which will help understanding
conversations and identifying opinion leaders within social networks.
In the digital age,data is the raw material from which one can get all the information about online users. But knowing how to collect, analyze and draw conclusions from this often unlimited source of information, it is sti
ll something equivalent to, let’s say, buffalo hunting for lions: what matters today is no longer where to find the information but being able to tear the prey to digestible pieces once you have it in front.
Part of the Intelligence team efforts are focused on this direction and the algorithm that has just been designed is just another cog of the gear that will lead us to better understand people and give them what
they need. Among other things, this algorithm will automate the
identification process of topics exclusively talked about in social media: something crucial in light of the flood of information generated every day. Tracking a conversation between seven people on Youtube, for example, could be manually done. But when the number of speakers are in the thousands and the channel is not only one but hundreds – whether they
are blogs, forums, microblogs or any other social platform – there is too much noise to hear anything.
Taking our analysis to the extreme: Can you imagine knowing what brand of shampoo teenage girls like in the very
neighborhood of Tribeca in Manhattan? Or who’d like to buy the jacket George Clooney wears in his latest movie? Although it
still may sound like a trip to the future, this is where the efforts around topics detection are heading towards.
Furthermore, this algorithm will also help tracing the path of a message to find opinion leaders. By following the
spread of a message until it reaches its source, you can know who originated a conversation, whom some people follow in
his or her reflections, or who is considered an expert on a specific topic. It is all about being able to find the MissPandora of
perfumes, laptops or automated pop music and, most importantly, in real time.
In addition, when investing online, having the ability to analyze data is key. Online advertising has a high degree
of inefficiency due to -among other reasons- we still find ad campaigns that are limited to replicating the offline
model. Future success, however, will be for those who can guarantee a completely customized campaign, offering
thoroughly personalized ads for each and every person behind a screen. And this is still a fairly unexploited field for both
Google and the smallest company in the sector.
Despite the fact that there already are some initiatives in the market in order to sort the noise generated in social platformssuch as Twitlogic-the majority of them present several shortcomings;
To mention a few : A)there are other Social Media platforms beyond Twitter but the majority of algorithms on the market still focus only on
the main ones.
B) there is still a long way to go in the field of reading comprehension in Social Media. With just 15 minutes on Facebook
one can realize that people could be talking in Esperanto and it wouldn’t make much difference: language in social networks
is full of abbreviations, misspellings and slang uninterpretable by machines. This algorithm also considers the context of the
words to detect whether the “blackberry” that we are talking about refers to mobile phones or recipes.
