I ran into an interesting edge case on the Chapter splitting. I ran the script on a theological book that included many references to the Bible, often saying things like "In Romans chapter 3" or "in Genesis chapter 6" or "in the third chapter of Ezekiel".

I ended up with 185 or so chapters, some only a few seconds long.

I wondered about finding a way to isolate those chapters (maybe adding some logic that excluded a list of words, in this case the names of the books of the Bible" either before or after each word of chapter; though that may not be perfect, if a chapter ends with one of those words, for example).

@prodigalstatius this is an excellent example of how to expand the "excluded phrases" list; this is essentially a list of words (or "phrases", if you will) that, when grouped together, the script detects via pattern matching and ignores during chapter splitting.

If you don't mind and have the time, please compile a list of all (or as many as you can) of these phrases that triggered chapter breaks and submit them in a GitHub issue. From there, I can take them, add them to this list, and it should reduce the number of erroneous chapter breaks. This would ultimately help everyone using this script in the future!

I am traveling at the moment but I may be able to do that this weekend.

This is such a great project. I've used it a few times.

I do have a book I want to chapterize that is going to prove a challenge, though. Grimms Fairy Tales are all individual stories but none start with Chapter at the beginning.

I ran into an interesting edge case on the Chapter splitting. I ran the script on a theological book that included many references to the Bible, often saying things like "In Romans chapter 3" or "in Genesis chapter 6" or "in the third chapter of Ezekiel".

I ended up with 185 or so chapters, some only a few seconds long.

I wondered about finding a way to isolate those chapters (maybe adding some logic that excluded a list of words, in this case the names of the books of the Bible" either before or after each word of chapter; though that may not be perfect, if a chapter ends with one of those words, for example).

This is why I love both language and programming! It’s so complex.
For instance, the code already has “Matthew Chapter” on the avoid list. However, it wouldn’t make sense to put all 66 books on the avoid list, as it’s really common to simply refer to the chapter and verse if you’ve already mentioned the book (I.e. “In John chapter three … and in chapter four, Jesus goes down to Samaria…”). This makes it anything but easy to figure out what would actually be a chapter division.

I previously found in the code where the list of division keywords were, and simply added “tale” to my own local copy (I’m trying to split Beatrix Potter for my son). However, the second book (Squirrel Nutkin) starts off with the line, “This is a tale about a tail.” The AI put a division on both. It’s amazing to see how important context is!

I wonder if you would be able to expose the chapter divisions after the AI finds them, such as in a file? I’m not sure if they’re shared as timestamps or something else, but if we could have access to the list before the program splits the mp3, we could manually adjust for all the erroneous (and really difficult) divisions…