Contractly | Devpost

Home (w/ download for chrome extension)
Contract AI Tool
Contract Analysis Info
Contractly Privacy
Contractly Privacy Extension In Use

Inspiration

How many words do you think the average privacy policy contract has? 500? 1000? The answer is 4000 words. That's about 16 minutes of your day. And how about contracts and legal documents? A 10-page agreement could take even an expert hours to dissect. Thus, we are excited to introduce Contractly.

In an age of contracts, safeguarding your personal data and proofreading agreements is crucial. Discover the significance of website privacy: from protecting your information to building trust, staying compliant with laws, and safeguarding your reputation. Explore the world of online safety in addition to contract analysis and make informed choices for a secure life.

What it does

Discover the transformative power of Contractly AI that has the potential to revolutionize the way businesses manage and analyze contracts. Contractly AI employs artificial intelligence to enhance contract management processes, and increase efficiency, accuracy, and time savings. By automating tedious tasks, extracting key contract data, and generating valuable insights, Contractly AI allows businesses to optimize contract lifecycles, minimize risks, and unlock greater levels of productivity. Say goodbye to manual contract review and embrace a future where Contractly AI is your trusted ally in contract management.

Our next notable feature is our privacy chrome extension tool that enables the average consumer to quickly understand and digest privacy statements. Whenever you visit a new webpage domain, the extension will automatically poll our API that provides a score and a summary of the privacy policy. With a click, get a clear, concise summary of a website's privacy policy, saving you time and ensuring you understand how your data is handled. Empower yourself with simplified privacy information for smarter online decisions.

How we built it

We built Contractly AI using a Large Language NLP model trained on a dataset of over 500 contracts spanning NDAs, mergers, acquisitions, and more. With an expertly-annotated 41 questions powering this Question Answering Transformer model, Contractly is able to extract important answers to the crucial questions annotated by sector experts.

We designed our front-end in React and Tailwind, and our back-end is in Python and uses Flask, MongoDB, and Langchain. We focused on ease of use, emphasizing simple UI and UX design in the front and improving our processing pipeline in the backend allowing individuals and businesses alike easily upload contracts and receive the analysis ASAP.

Challenges we ran into

Throughout the design and development process, we were faced with an array of challenges. Perhaps the greatest was integrating the entire application into one full-stack web app. In order to integrate the model, we faced the difficulty of creating a pipeline for processing user input into a readable text format for tokenization and feature extraction. We also had to account for incorrect logic that might arise in our NLP model such as extracted substrings having end logits come before starting logits or overcoming the token limit. Additionally, dealing with scraping webpages as well as chrome extension permissions was difficult to handle.

Accomplishments that we're proud of

A 10-page contract can, on average, take even an expert almost 2-3 hours to throughly analyze and understand. With Contractly AI, a 15-page contract can be analyzed in under 15 minutes—a drastic difference.

For our Chrome Extension, we were able to bring the time required for analyzing privacy statements with thousands of words in seconds. We also successfully integrated MongoDB into our extension, allowing for quick, <1 second retrieval of summaries and scores that had been previously been stored by our backend.

What we learned

While we had experience and played to our strengths, we purposely designed our project so that it would allow us to explore. breadth of new topics—from entirely new frameworks like Flask to actually deploying a ML model with a working full stack application to even developing a chrome extension.

This project also enabled us to dig deep into Machine Learning and preprocessing, requiring us to build a pipeline that processed text and tokenized substrings for prediction.

What's next for Contractly

First and foremost, our primary goal is to deploy our full-stack application to the cloud for production use. For our actual application, our next steps are to integrate our privacy model into mobile applications as well and continuously build upon our model to decrease prediction time. We would also like to enhance our data sets for our contract to provide a wider breadth of analysis in the future as well. Contractly has many applications across a multitude of sectors, such as business and law.