We all get too many emails. And writing out professional responses for every email can take a lot of time. Templates are lame, but bots are AWESOME! So we thought of how a bot could solve endless email syndrome!

What it does

MAILBOT reads all the emails in your inbox and summarizes the most important ones into short summaries that are sent via a Spark Bot. You can then speak to your MAILBOT as you would a personal assistant with a short description of what you require and MAILBOT will create professionally formatted emails so that you don't have to spend time answering endless emails!

How we built it

We used Node-Red on IBM Bluemix to build 3 Node.js Servers with JavaScript. Server 1 connected to Office 365 to read and send email messages. Server 2 connected to IBM Watson Alchemy to summarize emails and provided Speech-to-Text from Watson. Server 3 accepted webhooks from Cisco Spark and posted messages back to Cisco spark as the bot.

The Harmon headphones connect to a native iOS application that streams the audio information from the microphones up to the server in order to convert the speech to text. The app was built in XCode using Objective-C programming.

Challenges we ran into

We experienced out of memory server crashing due to sending streaming audio to our server. Data structuring was an issue, specifically data parsing for Watson caused us some issues. Luckily, we pulled it off in the end! Thanks to the amazing Watson mentors :)

Accomplishments that we're proud of

We're proud of the fact that half of our team is female devs! We're proud of implementing so many APIs and are really excited by how well they worked together!

What we learned

We learned that Natural Language Classification is VERY difficult. We also learned how to scale speech to text applications on IBM's Bluemix infrastructure.

What's next for

We'd love to scale this product out into a more complex project and will likely continue building this product as we love building super cool A.I. Driven chatbots!

Built With

Share this project: