People often ignore the important details of their contracts, whether it's a house lease, or terms of agreements of a service they are using. Sometimes, missing such important details might inevitably lead to signing without being aware of how 'shady' the contracts are, and they fall into a trap that they'll struggle to climb out of for a long while.
What it does
Our project uses machine learning to define and detect potentially 'shady' statements from the contracts and provides detailed report for the user, after which the user can choose to either electronically sign or don't sign the contracts based on the conclusion the user draws from reviewing the 'shady statements.'
How we built it
We used machine learning algorithm to conclude which statements are considered 'shady' and built a webapp using Vue.js for the frontend and Python and Flask for the backend. Lastly, we utilized the Docusign eSignature API in order to email the user the contract for the user to provide an electronic signature.
Challenges we ran into
We struggled the most with putting different components of the project together. As a team of first time hackers, we found ourselves spending a lot of time simply learning and figuring out the dynamics of hackathons as we developed our project. It was very difficult learning an entire framework as none of us have ever developed a web app from start to finish.
Accomplishments that we're proud of
We had many proud moments and many frustrating moments during our time at Calhacks this year, but some of our proudest accomplishments include: successfully detecting 'shady' contract statements, getting the web app (made using Vue) to resemble the wireframe we had made, and implementing anchors to create signing tabs with Docusign API.
What we learned
We learned not only what hackathon is like and why Yerba Mate is an amazing drink, but also how to implement machine learning, use APIs and Node-based frameworks, and how to work with passion and lack of sleep as a group.
What's next for DocuScan
First we will refine the end-product that we made during our limited time, then we see great potential to scale the project to include different types of contracts, since we only have training model for terms of agreement for now. Then we expect to use similar concepts of 'shady' detection to try and tackle the problem of undesirable employment offers that a job-seeker might receive from different companies. We also did not have time to include pdf text recognition and the ability to add multiple files at a time.