Lawyers need to anonymise text in legal documents while sharing it in the public domain. Their existing set of tools didn't cover all the named entities in the document and required manual intervention by a lawyer, wasting precious time and resources. Since the data privacy of the users is the utmost priority of any legal firm, we decided to help them out so that their relationships with their customers could be long lasting and trust worthy.
What it does
Our tool analyzes e-mails using Named Entity Recognition. This identifies references to various entities, such as people, locations, and organizations. These references are scraped and replaced by a set of pseudonyms, such that the same entity is referred to by the same name and it all still makes sense. Legally Anonymous maintains a database of these entities and their references across the mailbox.
How we built it
We built a webapp, based on Python, Flask and the NLP framework SpaCy in the backend, and React in the frontend.
Challenges we ran into
Coordination in a hybrid-remote team, domain knowledge in the legal field.
Accomplishments that we're proud of
A clean and beautiful UI, working on a real-world use case, a moderately funny video.
What I learned
That LegalTech is a thing and in fact offers some interesting challenges and lots of potential.
What's next for Legally Anonymous
NLP and LegalTech seems like a match - we could imagine building more of this in the future.