jonathanstray

Thanks for this. I have updated the broken links.

While I do appreciate the role of a proper citation, I'm reluctant to do it for blog posts (as opposed to my actual papers) both because it's time consuming and because for most people, links are a more natural and readable format.

Nice! An excellent baseline, something to beat.

I didn’t try separating out tables because the total field isn’t actually “inside” the table in many cases. Certainly the other fields I want are not.

pdfplumber seems mostly ok at extracting tokens. Sometimes it seems to combine tokens that should be separate. I suspect a few percent of the error is actually problems earlier in the data pipeline, as opposed to the model proper.

It’s not a binary classifier. Every invoice in the dataset has a total amount written somewhere on it. Accuracy here is whether the network chooses the correct token from each PDF.

Heh. It’s my work, so maybe I can clarify the goals. This is meant to be a proof of concept. I took a week and was able to show that relatively simple deep learning techniques are capable of generalizing over unseen form types with high accuracy. I also showed that tokens-plus-geometry is a viable format, and that hand-crafted feature engineering is still necessary (and still used in SOTA approaches).

I also believe that preparing and cleaning this data set, and bringing a challenging investigative journalism problem to the attention of other researchers, would be valuable even if I hadn’t done any work on this baseline solution. This is a problem that journalists currently expend a huge amount of time and money on, which reduces the effectiveness of transparency around political ad spending information.

God no, don't do it in git. Otherwise no one outside of developers will ever look at it, and the whole point (as I understand it) is to bridge the tech, law, and politics worlds. So some of the primary users aren't developers and have probably never seen a command line.