Inspiration
Current legal drafting tools implement rule-based approaches to create documents. They follow standard templates and are difficult to customize as per the organization's needs. This results in time-consuming reviews and negotiations between the client and the organization. A legal personnel takes anywhere from 30 mins to 2 hours to review an NDA depending on the experience. We believe we can reduce this to less than 10 mins.
What it does
With its advanced algorithms and natural language processing capabilities, our AI-powered legal document solution can draft high-quality legal documents, suggest relevant clauses, and make edits with unparalleled accuracy and speed. Say goodbye to manual processes and hello to increased efficiency, productivity, and profitability. We make use of MS Word add-in technology so legal teams won't have to migrate to a new software.
How we built it
We use various NLP algorithms to classify documents, generate clause suggestions, summarize paragraphs, and extract entities. Specifically, we made use of HuggingFace and OpenAI libraries to accelerate the development and deployment process by re-using existing pre-trained models. We also make use of Microsoft word add-in technology to bring functionality to the data. This helps legal teams in quickly adapting to our product within a familiar software environment. We use ReactJS to build the MS Word add-in UI.
Challenges we ran into
MS Word add-in was a new technology for us and it required a steep learning curve during the hackathon. We also needed to take extra care in processing text coming from word documents as it sometimes contains special characters that are hidden. These hidden characters caused challenges in running NLP algorithms and performing operations such as search and replace using MS Word add-in.
Accomplishments that we're proud of
We successfully deployed the MS Word add-in built using ReactJS by following the standard practices mentioned in the documentation. We also optimized the deep-learning models to analyze the documents within 1 second. All of these points add up to provide a great user experience.
What we learned
We learned how to build MS Word add-in using ReactJS. We also learned how to implement Text Generation and Text Summarization capabilites using HuggingFace and OpenAI libraries in Python backend using Flask framework.
What's next for SALT: Smart Assistant for Legal Tasks
Expanding capabilities to other legal documents such as Master Service Agreements, Independent Contractor Agreement, and Statement of Work. Enhancing the drafting capabilities of the deep learning models using fine-tuning techniques. Addition of chatbot to the user interface to enable users to perform custom operations.
Built With
- deep-learning
- flask
- huggingface
- javascript
- openai
- python
- react
- word
Log in or sign up for Devpost to join the conversation.