This is an international legal dictionary, an experiment in improving the situation: glossaries are scraped and parses from official sources: https://github.com/public-law/open-gov-crawlers. The results are saved as datasets in well formed JSON with Dublin Core metadata: https://github.com/public-law/datasets
I add Library of Congress subject headings to the sources, to enable filtering (still to come).
The web app is basically an old-school mashup, which I've always liked.
Another experiment is using the Dale-Chall readability formula to improve the reader's experience. Here's an example of it at work:
https://www.public.law/dictionary/entries/amicus-curiae
This is an experiment, using readability as a relative metric. I.e., not extracing an absolute grade-level score as its normaly used. Instead, using it to compare different definitions of the same phrase. My theory is, there's strong scientific validity for this use, even when applied to very short passages: All I simply want is to figure out, "Which is more readable? Passage A or B?" And then, my code sorts the definitions in order of readability to (theoretically) produce a newspaper-article-like effect: A reader can read the first couple of sentences to get an overview of the story.