Emotions analysis for word

Recently I had the idea of creating a add-in for office outlook and word that would allow people to analyse the sentiment and emotions of the text they were writing. Not only as a tool people could use to improve their communication, but also as a way of crowd sourcing quality improvements by allowing people to report when they see an obvious error. Computers for example don’t get sarcasm or irony, because they just don’t have a good sense of humour.

I started out by using Vader sentiment analysis as the back-end. This worked but did not quite give me a rich enough context. Next step was to connect the add-in up to IBM’s Watson Tone analyser to give me the emotional tone in the text.

This worked great too, but the IBM Tone analyser has some short-coming’s, but this is a discussion for another day. I then thought, given I has some time during lockdown and a keyboard, could I cobble together my own engine to analyse the emotional tone in text. This took me on a bit off a journey (and down a couple of rabbit holes), looking at how sentiment analysis in general works and emotions can be identified.

So today I have a beta version of an engine that analyses emotions and tones connected to an office add-in for word. Theoretically, it can identify up to 32 different emotions(have not tested this, yet) in a given piece of text. At the moment I am focusing on the 4 base emotions anger, fear, joy, sadness and trying to improve the accuracy of the engine for these.

For my linkedIn connections out there, I am looking for a few people to help me out with some testing and feeding back. DM if you can. I am also interested in peoples ideas of how you could improve or use the detection of emotions.