Legal Loom

Note

This project is incomplete and is not intended to be a submission for any of the sponsored projects.

Inspiration

Our group was inspired by Morgan & Morgan's sponsored problem statement. We think that generative AI is the strongest solution to assist the virtual legal industry. Not to mention, by attending Microsoft's Semantic-Kernel workshop, we all were inspired to implement that as our backend for generative AI.

What it does

Legal Loom serves as a web-based case summarizer, utilizing OpenAI's GPT 3.5 model to provide summaries of given input. By providing the text from sms messages, reports, and other media, it will provide summarized feedback of the case.

How we built it

We built this project around semantic-kernel. It was implemented as a backend using Python, and connected to a frontend using Axios. Thus, the rest of the project was implemented via React, and deployed on the web using Vercel.

Challenges we ran into

Overall, this was out team's first hackathon. We are not strong in knowledge with React or web-development, so learning this, along with implementing new technologies, proved to be difficult. Not to mention, implementing semantic-kernel, which thrives with Azure which is another skill outside our understanding, was difficult.

Accomplishments that we're proud of

We are proud of the fact that we now understand the basic principles of React and how to make an application such as this. Not to mention, we have knowledge of how to implement API keys for GPT's generative AI models.

What's next for Legal Loom

Firstly, this project will be polished in the sense that it will properly display output to the user after the prompt is given. From there, it is a matter of formatting the application to fit the original requirements strongly. It can take files as input, and send them to be processed either simultaneously or at separate times - which also requires the successful implementation of semantic-kernel's memory system. Then, it will be able to provide output in the requested form, whether that be summaries, parameters, case numbers, or others.