Sunday, 9 November 2014

You can build a lot in a day

On Saturday at Brandwatch HQ, we hosted our inaugural student event, which we called BestIn.class. When I was an undergraduate student, I was often intimidated by industry, especially when reading and hearing about interview horror stories; they gave me the same feeling I used to get worrying about exams. We thought it would be a good idea to open up the office to university students in the area (Sussex undergraduates were the biggest attendees) for a day of coding with a show and tell session at the end. The premise was that the next generation of programmers could come and see that we aren't so scary after all, and that much of the code that they are writing from day to day isn't all that different from what we're writing from day to day.

I was hesitant going into the event, as I was primarily concerned that people wouldn't turn up. I was wrong there. I was also concerned that 6 hours wasn't enough time for people to build anything cool. I was wrong there also. We gave the attendees a number of code samples ahead of time so that they could prepare. These ranged from code to connect to and parse the Twitter stream with Hosebird to a Spring Boot application that could login and access your friends on Facebook. However, we kept the datasets secret until the day to test everyone's creativity. We chose a dataset of 2500+ news articles on the day of the World Cup 2014 final, another with 5000 posts from the last week about ebola, and another with 5000 posts around the week of Robin Williams' death.

What our attendees came up with was extremely impressive:

  • An attempt to automatically classify tweet locations when no geo-coordinates are given (a problem we struggle with ourselves)
  • An implementation of a Caesar cypher for tweets
  • A map plot of the most talked about African countries in the ebola dataset
  • A map plot of where the ebola articles were posted from
  • A live Twitter stream with interactive cat GIFs driven by the emotion of the posts (a rules-based classifier)
  • A search engine for sentiment in different countries for World Cup players using Apache Lucene
  • A sentiment graph over time for the articles on the World Cup finals day
  • A 3D word mesh of the common words and their links in the posts around Robin Williams' death
Not only were the projects impressive; the enthusiasm for the event was extremely high, with attendees fighting to get everything finished before the presentations, and a real urge to show off what had been worked on. In short, I was extremely impressed, and I'm really looking forward to hosting another event like this in the future. Thanks folks!

No comments:

Post a Comment