Howdy, howdy, howdy!
As is our inimitable style this instalment is a little late. This is because here at 23Squared HQ we’ve been thinking about what to write about! We’ve had lots of ideas knocking around but have finally settled on a plan.
This blog is going to be a short one, a mere introduction to the fun to come. This is because we’re planning to do a project over a few blog posts, so we thought we’d spend some time introducing the idea.
There’s been a lot of media attention lately about online trolling and abuse. This week Twitter has released new features to address the problems that its service suffers from. This got us to thinking, is there a way of automatically assessing a piece of text to ascertain it’s theme and/or content.
Well we think so. We’re going to be looking into the Standford NLP library to see if we can achieve a successful implementation of sentiment analysis. As we mentioned before we’re going to be breaking this down into a few blogs and build the system step-by-step.
Initially we’re aiming to build something that can classify a piece of text into overall negative sentiment or overall positive sentiment. From there we can move on to more complicated analysis to decide whether or not the text is offensive or not. We appreciate that the caveat here is that being offended by something is a very personal thing, so we’ll have to do some benchmarking on the assessment to make the decision based on what most people would find offensive (again, caveats apply!).
We’re to play this project by ear and let it evolve as we go so it might not be the most coherent journey, but it’s going to be a riot nonetheless!
Tune in soon for the first technical instalment of this blog series viewers!