KeyError: ('___BEGIN__', '___BEGIN__') about markovify HOT 7 CLOSED

Usually means there was some kind of symbol it doesn't support (parenthesis for example). Probably a duplicate of #84 .

from markovify.

Thanks for pointing to #84, @kade-robertson! That does seem to be a likely explanation.

The issue, however, is typically less that the corpus contains an unsupported character, and more that all sentences contain such a character. But it's a bit difficult to diagnose definitively without knowing more about the corpus. Could you share the corpus you're using @icyphox, or a sample of it?

from markovify.

I'm actually generating the corpus from a user's Twitter feed, and that's most likely it, as @kade-robertson said.

from markovify.

I've made #93 as a partial workaround to the issue, you can disable the rejection regex entirely with well_formed=False in Text initialization, or provide a custom regex with reject_reg to substitute. This may not be ideal as you will then get unmatched parenthesis in the output. There's no one good way to solve the issue unfortunately -- You either don't use that text at all for learning purposes, or you enable unmatched parenthesis or weird punctuation (which you could potentially handle later, or make a different kind of Text which will interpret text in parenthesis as entirely different sentences).

from markovify.

@kade-robertson I am experiencing this problem. I see that adding well_formed=False was added but for some reason I don't see that option in the main branch anymore?

from markovify.

@crypdick @icyphox #93 was just recently merged in if you are still trying to use this library with your original corpus. :)

from markovify.

Oh that’s awesome! Although my Twitter bot is long dead. :D
Guess I’ll close this now.

from markovify.