Well, after about a year of inactivity (partially my fault, partially not) I’m back to the Roget 22 project. If you don’t remember what this is about, I can hardly blame you; here’s the post where I explained things .

I’m not going to make any big predictions or lay out a lot in the way of plans right now, but I did just re-read my last batch up devlog entries, and some things have changed from my thinking at that point. So here’s a list of what’s happening:

  • No more Swift. That was just me trying to find more use-cases for Swift, so that I could get a better handle on it as a language
  • New source material. Public domain has moved on, and so is the project. I’ve ordered a 1930 Roget’s from eBay to use as the scanning/OCR source
  • I have run scanning/OCR tests. It won’t be super-fast, but it also won’t take as long as I had worried it might
  • Also, development will not be waiting on the scanning effort to be complete. They will start together
  • Recent experience with Thrasher (whose development I should start detailing here) has gotten me more comfortable with Javascript work. There might not be a “placeholder” CLI client this time around; work might start on a client and server immediately and concomitantly with the processing toolchain
  • I believe there is a human-readable and machine-parsable way to implement references, which won’t break if/when headwords or senses get reorganized (it will still require refs to be resolved on a second/nonprimary pass of the source, but I think this is always going to be the case if you’re automating things)

I plan to get started as soon as I have the new text in hand, and may smooth out a few things in the scanning process while I’m waiting.