They Say

About six years ago, I looked at starting a company in the area of, what has now come to be known as  'sentiment analysis' - that is the ability to measure the buzz of a brand, product or even a political topic online, right across the web through an automated language recognition system. We even had a name for it, 'Flux' - isn't that always the easiest part of getting a business off the ground. But what we soon discovered was that language interpretation software just wasn't up to it - for example, when does 'bad' mean good and when does it actually mean bad? But recently I have seen publicity about a company called They Say - it was featured this morning on The Today programme on Radio 4 and it sounds like they may now have the language problem cracked. It helps that it's an Oxford University spin out company, coming out of something called the Isis Software Incubator and according to this morning's interview their Computational Linguistics Software has the ability to determine when for example the mention of 'bacteria' might be good - in a yoghurt,  and when it might be bad - in spreading a virus. This is what they say about themselves:


Our software performs exhaustive fine-grained analysis, resulting in accurate human-like sentiment reasoning. Our approach of employing linguistic intelligence to complement machine learning techniques enables new levels of insight and transparency, while providing best-in-class results on key performance metrics (independently verified by exhaustive third-party studies).


It will be interesting to see how they get on but there aren't too many brands out there who won't want an accurate interpretation of their online sentiment across blogs, Facebook, forums, customer generated reviews, et al