Building software that 'thinks' like a lawyer - Ver 2

Last year we wrote about some challenges of automating real estate due diligence, focussing on optical character recognition (OCR). If presented with the images from that article, a human could quickly assess what parts are included or excluded, where a machine learning model may struggle. But what about the opposite: where a model can easily identify something a human cannot?

Take a look at this single page from a 43-page lease of a flat:

Without the helpful blue box, how easily do you think you could confirm if your beloved pet was allowed to reside with you? With enough time, effort, and paracetamol, you might find the key clause referring to “bird reptile dog” that would give you a clue. It would be understandable if you missed it, unless you were acting for someone whose canary was excited about moving.

Technology has a role to play in delivering critical information like this to people who can make a judgement on it. Once we ran this lease through our OCR tools, our model predicted with 79% certainty that this clause was a restriction on keeping pets in this premises.

Why not 100%? Well read the full clause and judge for yourself! Lawyers and insurers get paid for their valuable experience in assessing whether such issues are likely to be a problem. But the technology can accelerate the process and help to avoid missing such critical pieces of information altogether.

We’re hiring

If you are interested in helping us automate real estate due diligence (and making sure pets have a place to live), please see our open roles and get in touch with us via our Careers Page. If nothing quite matches your experience then still feel free to connect and message me, Neill Pemberton, directly on LinkedIn and I’d be happy to have a casual chat.