Ask HN: What would you do with 1.3M laws?

2 points·a3camero··6 comments

I'm the CTO of a law search engine that has 1.3 million laws from 47 countries. We're now wondering what else to do with our giant database of laws. What kind of text analysis could you run? ML algorithms? Clustering? What would be interesting for you to see as an experiment?

View on HN

Depends how complete the corpus is, it would be interesting to compare language across laws in different countries to see if how similar they are and possibly detect the influence of lobbyists working on an international level.

We translate them into English using Microsoft/Google Translate (which is what you need to do in order to compare between languages) so the meanings may not be quite so easy to compare (because it relies on the current state of the art of machine translation of language - still not at professional legal translator level).

That's a really interesting point about detecting lobbyists, although my first thought is that most language similarities will be due to international treaties.

I'd also be interested in seeing which jurisdictions have copied from each other. Maybe there are "leaders" in developing laws.

As for completeness: it relies purely on laws that are published online, which is always not the complete corpus, even for advanced countries.

These laws often make reference to other laws, right? I'd like to see the things referenced seamlessly merged in.

Also, it would be awesome if they could be presented in such a way that laymen could understand what laws actually mean. Legal meanings of words are often very different than the common meanings.

We've actually done some work about extracting definitions from laws. But the meaning is still providing using legal language so they may not help too much for the lay person.

Reply on news.ycombinator.com