Proof of concept: parsing PDF tree felling permits

[electricmonk]

Hi
Following #587, here's a proof of concept that reads from the Tel Aviv city hall website, downloads PDFs and then attempts to parse them.

Currently the results are so-so; only about 90% of the PDFs are text - the rest are scanned images, which would require integrating OCR. Even those that are text, are not deterministically parsed and would require more work to improve the ability to extract reliable data. Currently about 70% of data is parsed, but I'm scoring all fields equally.

The important question before I move forward would be - is this better than nothing? should I invest more time?

[CLAassistant]

All committers have signed the CLA.

[gruppin]

hi, unfortunately this is not better than nothing.
our integrity for our users is to have all the tree licenses, not only part, but they relay on us to notify about every tree license.
so in this sense, notify about partial set of licenses is even worse than not notifying at all, since we might mislead our users with a false image of reality.
I don't think you should invest more time in it.

[electricmonk]

Although, we can extract 100% of street addresses from the PDF file name, and we can conceivably create a unique id from street address city and publication date. So we won't miss any petition - just have missing data for 10%-30% of them. Wdyt?