Too many corporate executives waste their time in hourly meetings, when they could simply get the value of that meeting out of reading a 3 minute summary. We wanted to address this problem by providing an accurate and inexpensive way for everyone to get the most out of their meeting's minutes.
What it does:
Our hack uses Kinect Voice and Beam angle recognition to take in details of the vocals of a meeting. Using FTP technology, that recording is then sent to a web server which runs a Microsoft Oxford-based transcription algorithim that converts the speech into text. The raw text then branches off into two paths: The first takes live transcript files from parts to the meeting and outputs them in real-time into a OneNote notebook, which can be shared with everyone involved in the meeting, even though they may not be physially there. The data also travels through a self-made natural language processing algorithm, to extract the important highlights of the meeting. After it is processed, the summary is stored in a MongoDB database. From there the summary is posted to the same OneNote file for each of the meeting members to edit, and review, while keeping the original audio and transcript on record. The meeting is also automatically posted to the Google Calendars of involved employees for easy future reference.
How I built it
We built the Meeting Minutes program using a wide variety of software programs, languages, and operating systems. The sound and relative location data comes straight from the Microsoft Kinect, through the Kinect SDK. Using Visual Studio on Windows OS in C++, the audio is saved and split into smaller segments that are fed through a FTP server powered by Python to the voice processing, transcription, and NLP hub on MacOS. After the data is processed and ready for publishing, it is transferred using a MongoDB database an Ubuntu machine running a Java applet that uses the OneNote API to transfer the data into notebooks. Finally, the Google Calendar segment of the program ties it all together by automatically adding new meetings to the calendar of everyone involved, and places a link to the OneNote notebook.
Challenges I ran into
The biggest challenge was probably the compatibility issues across all of the API, OS, and programming language implementations in our project. Due to the limited functionalities of certain APIs in certain languages, converting all of our data and passing it around through a long processing chain was very difficult. We also ran into signnificant technical difficulties invovling a teammate's computer, which we had to navigate around to make our solution work.
Accomplishments that I'm proud of
We are really proud of how well the team came together to take on our programming challenge, given that we represent 3 different universities, and have never worked together before. We celebrated every connection that we managed to create in our long link a real-world need in inefficient minutes, and
What I learned
It's really surprising how much a team of well-rounded individuals with distinct strengths and weaknesses can accomplish compared to one person alone, even if they were given 10x the time. It is really astounding to see the amount of resources that technology companies provide to developers, and it only takes an idea and a little bit of initiative to create something amazing.