I haven't actually tested this, but I wanted to mention a couple things.

Keito said:

IT ONLY WORKS FOR IMAGES THAT ARE NAMED AFTER THEIR MD5 CHECKSUM. NO TAGS FOR YOUR "yomamma.jpg" IMAGES!

The md5 is hard to rely on for this (particularly for images that have added metadata), but why not attempt to calculate md5 when the filename match fails?

I know certUtil -hashfile somefilename MD5 works for a single file, but you might be able to push your input through that and get something meaningful out of it (I haven't tried doing this). Else I'm sure there's a freeware utility out there that can do this for you.

Secondly, I highly recommend using our API to retrieve the tag information. It should make queries run noticeably faster but mostly it will be much much easier for you to parse and less likely to break than trying to scan the page for strings.

I prefer JSON myself, but we support XML as well.

JSON: https://e621.net/post/tags.json?md5=b727f5f7982e8d6d1c6e01754562c075
XML: https://e621.net/post/tags.xml?md5=b727f5f7982e8d6d1c6e01754562c075

Just note that the tag types are actually written incorrectly on the API help page (I haven't gotten around to updating it yet). These are the correct ones:

"General"   => 0

"Artist"    => 1

"Copyright" => 3

"Character" => 4

"Species"   => 5