I was recently able to make good progress on a side project that I’ve been working on for some time. The idea is to have a ‘virtual assistant’ that will find action and meeting intents from the different channels of communication I’m using daily, for work or personally.
Being always remote means that the digital communication has increased exponentially, and it can be overwhelming to keep tabs on everything that is going on. The virtual assistant is hopefully going to fill the gap I’ve left in my emails, chats, and comments.
The idea is to use a language model to identify intents from communication, it appears to be working pretty well so far, even without doing any fine-tuning.
Below are some examples I captured from the demo I’ve put together.
As you can see in the examples above, the tool highlights the sentences that match one of the defined intents (action, meeting, or task completion) and displays all the sentences with a confidence score above 2.
If you are interested in trying it for yourself, I have a page up and running in my lab: http://metah.ch/up/home.html I’d love to get your feedback and if you notice anything that is not correctly inferred.
Next step
Now that I have the end-to-end working, my goal is to enable the functionality directly into some email program (like Gmail or Outlook) using add-ons.
If you would like to stay informed about the progress, please share your details below: