Email Parser Screenshot
Image to Text
Everyone in this group has missed out on killer campus events whether free food, trips to NYC, or spontaneous snowball fights. It's just impossible to keep up with Dartmouth's endless clubs, student organizations, and hyperactive listservs. We wanted to give users a chance to access all the happenings on campus without frying our brains with 128 unread emails.
What it does
Our project uses an Outlook extension to parse data from the many blitzes that go out to campus. Outlook Web currently has a “suggested meetings” feature, but it often doesn’t pick up the correct dates or pick up that any event is happening at all. Not to rag on Microsoft’s natural language processing, but at least on our campus, blitzes are written in a very specific style. Our program is able to get past the fluff for information in a quick, digestible format.
Eventually, we will output this data to an Outlook calendar of suggested events where a user could access the ones they were interested in. We also built a front-end mobile application that allows users to log in and select a time slot to view suggested events in their calendar.
How I built it
On the backend, we created a contextual Outlook add-in (an Outlook extension) and built our own rule-based text parser to glean the event information from the emails. We also used machine learning to turn our many digital campus fliers into readable text even though we were not able to incorporate into our final product.
On the front end, we built an application for users to look at time slots in their Outlook calendars for something to do. We still need help to get the data we parsed on the email into a user’s Outlook calendar, but that would require a more extensive knowledge of Microsoft API.
Challenges I ran into
We were working with an API that we knew very little about and had to figure out the best implementation of our project without maintaining or storing user data. There were a lot of moving parts to this project and it was difficult to implement as much as we could without having a single streamlined product.
Accomplishments that I'm proud of
We know our product has a market. Two of our group members independently came up with this idea to solve their own issues with Dartmouth emails. However, the complexity of the problem required external help. We reached out to one of the Microsoft representatives to learn about the feasibility of our project and she got us in touch with some product managers she knew. With a couple of links to API documentation, we built our product from the ground up, going further than we expected.
What I learned
This was completely new tech for most of us, so we were really looking at how to best optimize our skills and our time. We had a pretty high-level concept and we had to combine a lot of separate programs into a single concept.
What's next for DoSomething
The name of the game is reducing email clutter. We'd like to improve our accuracy to the point that users would feel comfortable with using our service as a primary source of campus information. It would be great to automatically move the leftover emails out of the inbox and into a separate folder. Later iterations of the DoSomething app would have more extensive methods to filter and categorize events on the calendar. The DoSomething app could also use predictive learning to suggest relevant future events.
Beyond Dartmouth, we'd like to hack Outlook's "Suggested Event" technology and sift through incoming emails to create a prospective calendar for all upcoming events. Instead of individual suggestions of what can be turned into an event, we'd rather have the chance to review it all at once.