good catch, I think the classifier saw the "distribution" in the sentence and assumed it was drug related. I could probably run each sentence against an ai model to classify things..

probably run it against something small like haiku and not cost me an arm and a leg.

I'm exposing the database that I use to build the site now as well.

https://pardonned.com/pardonned.db

I don't know what SPARQL is.. so no :)

are you asking me? if so.. I don't think I understand your question.

I'm thinking of exposing the sqlite db, and now possibly this JSON file as well.. I can take both during the build process and throw them up on CF. Great idea, thank you.

consider http://datasette.io/ as well

It's a play on Donald Trump, after watching a Liz Oyer video linking a very plausible pardon for sale scheme, I wanted to initially build a site that showcased pardons just by Trump, but I realized that would be partisan and not as useful.

Ah, OK. Sorry I didn't get it right away!

Interesting story, but I do not see her name in the list of pardons and commutations on the source website twice.

https://www.justice.gov/pardon/commutations-granted-presiden...

She only shows up here

https://www.justice.gov/pardon/clemency-grants-president-don...

The first time was under the name Adriana Shayota: https://www.justice.gov/usao-sdca/pr/federal-jury-convicts-s...

correct.

i am calculating it if it's available in the sentence details. If the sentence details don't have a fine or restitution then we can't calculate it.

Would love if you can track this more deeply and sort/filter/search through restitutions and fines. The ones you know about, that is.

Hmm, I see the issue.. The DOJ website lists all of Obama's as once, so I need to modify the parser.. https://www.justice.gov/pardon/pardons-granted-president-bar...

Compare that to the other list. https://www.justice.gov/pardon/clemency-recipients

That's probably intentional on the DOJ's part at this point.

not sure why you think it's intentional. But, created a github issue, and will work on that today/tonight.. yay GLM 5.1 :)

https://github.com/vidluther/pardonned/issues/23

I meant it's probably intentional that the data being represented differently on the DOJ's website, not your tracker website.

Stuff like this is very common. For example, at the start of Trump's second term, the whitehouse history page was changed to make democrat presidents look bad -

https://www.whitehouse.gov/about-the-white-house/the-white-h...

That link is entirely about the East Wing ballroom expansion. I don't see any criticism of anyone there.

Swipe through the "Major Events Timeline". It would be funny if it wasn't so sad how petty it is.

Good lord, this is so pathetic

>not sure why you think it's intentional

It's entirely on brand.

clinton-1 and clinton-2 are distinct. I think it's more likely collected differently. The people gathering data will change. Someone with different data standards worked there for a while.