AI / ML · NLP

Law Suggestion using ML

An NLP tool that takes an everyday description of a situation and points the person toward the legal references most likely to apply, lowering the barrier to understanding your rights.

Role: ML engineering
Client: Research & product
Timeline: Prototype
Year: 2023

Plain-language input, legal-reference output

A classic NLP pipeline: clean, vectorise, classify

Built to be explainable, not a black box

Overview

Most people do not know which law applies to their situation, and legal language is exactly the part they cannot parse. This project bridges that gap: describe what happened in ordinary words, and the model suggests the legal references most relevant to it.

The challenge

Everyday phrasing is messy and rarely uses legal terminology, so the model had to learn the mapping between how people describe problems and how the law categorises them, with enough transparency that a suggestion can be sanity-checked rather than blindly trusted.

Approach

I built a classic, explainable NLP pipeline in Python: text cleaning, vectorisation, and a trained classifier over labelled examples, deliberately favouring approaches whose behaviour can be inspected over an opaque end-to-end model for a domain this sensitive.

Outcome

The prototype demonstrates a genuinely useful pattern: turning a plain-language account into a shortlist of relevant references, as a starting point for someone to research further or take to a professional.