Accurate Text Analytics Solutions From Feedback Ferret

Accuracy for text analytics is vital

It appears that a French company has just launched a new text analytics tool that claims to have up to 80% accuracy in identifying sarcastic comments posted online (http://www.bbc.co.uk/news/technology-23160583). The company says its clients include the Home Office, EU Commission and Dubai Courts. I wonder why these companies didn’t just ask Feedback Ferret for our solution which has been able to identify sarcastic comments since 2002!

Interestingly, the article concerning this new tool also cites Simon Collister, who lectures in PR and social media at the London College of Communication. He says there is “no magic bullet” when it came to analytics that recognise tone. He also says: “These tools are often next to useless – in terms of understanding tone, sarcasm, it’s so dependent on context and human languages. It’s social media and what makes it interesting and fascinating is the social side – machines just can’t comprehend that side of things in my opinion.” Mr Collister added that “human interpretation was still vital”.

Yes, we fully agree that human interpretation of text is VERY vital. Which is why the Lexicons that power Feedback Ferret have been built over more than a decade from human observation of actual comments from all client records. In this way, we capture the real ways that customers express themselves, with all the twists and turns in the language that they use.

We are continuously enhancing and improving the Ferret Lexicons. Clients do not have to do any updates themselves – this is all part of the Feedback Ferret service. Any improvements to the Lexicons are applied to your current and historic data to continuously bring everything up to date.

Automated text analysis is not perfect. In reality, it is not cost effective to reach 100% levels of accuracy across large sets of data. However, the Feedback Ferret text analysis methodology is far more reliable and accurate than keyword analysis used by Natural Language Processing (NLP). In side-by-side comparisons, the Feedback Ferret engine has been proven to deliver results which are substantially better than NLP, both in terms of topic identification as well as accuracy of sentiment scores. This is important, as you need to be able to rely on what the reporting tool is telling you so that you can make robust business decisions.

Feedback Ferret typically achieves over 93% accuracy rates for contextual analysis, ensuring that you can rely on the results that are reported through our dashboards.