1. A horrible mess of sed and grep, etc., which finds the most common words in an HTML document that are not yet in the pronunciation file. I go through the results to add more words, and plan to rewrite the script in Python.

2. A short Python program that parses an HTML document with Beautiful Soup, then runs through the text substituting words, ignoring words in capitals (they’re usually acronyms) and optionally italicised ones. It adds naming dots to capitalised words that don’t begin a paragraph or follow one of “?”, “!”, “.”.

The naming dot thing is a hack, certainly, but it does make things easier to read. I think leaving heteronyms in Roman, while a little ugly, is also essential for readability. E.g. translating back from your Firefox extension’s version of the Shavian Wikipedia page gives “kingsley red” for “Kingsley Read”. I’ve found this sort of thing out along the way, as I’ve now read two and a half books through my system, starting with just a few words in the database and building up to the point where I now get many entire paragraphs in Shavian.

Thanks for starting this whole project and catching my attention. I didn’t know about Shavian before I saw your first Planet GNOME post, and believe it or not it makes a useful test case for a commercial project I’m working on.

I don’t know why WordPress has such a stupid bug.

Thanks for the email: I’d be interested to see what you’re working on. Some thoughts:

1) Spelling of “Shavian”: yes, you’re right, sorry
2) Naming dot: Not sure how to implement this from the Latin alphabet; we could just add it for any capitalised word not at the beginning of a sentence, but it’s a bit of a hack.
3) cmudict uses a strange Pittsburgh-based accent.
4) I have no good solution for homophones :(

I just thought I’d point out a few bugs in your system (I hope the characters come through here…).
- You’ve got “Shavian” spelled wrong! It should be “·

thanks for the tip!