It would be great to be able to optionally search over the entire text of all books.
Calibre now does this, which is great, but I'd much rather be using audiobookshelf than calibre.
Audiobooks could be included in this. This repo gets the full text from audiobooks as a srt (subtitles) file (for the use of finding chapter headings and splitting up audiobooks). There could be an optional setting to automatically make the same srt files for all the audiobooks of a library. Then when someone does a full text search, it could also match against these srt files, and when a match is found, could link to the relevant place(s) in an audiobook.
Or to get even more ability, it could use one of the free neural network models that create vectors from text (for both the text of books and for your search term) and match on the nearest vectors. This would be able to match related ideas even if they don't use the same words.
I know this seems like a pi in the sky idea, but I thought I would put it out here anyway. Here's hoping!
leeoniya
commented
8 months ago
Describe the feature/enhancement
It would be great to be able to optionally search over the entire text of all books.
Calibre now does this, which is great, but I'd much rather be using audiobookshelf than calibre.
Audiobooks could be included in this. This repo gets the full text from audiobooks as a srt (subtitles) file (for the use of finding chapter headings and splitting up audiobooks). There could be an optional setting to automatically make the same srt files for all the audiobooks of a library. Then when someone does a full text search, it could also match against these srt files, and when a match is found, could link to the relevant place(s) in an audiobook.
Matches could be done with fuzzy match model like a [Levenshtein-distance-based fuzzy search algorithm](https://github.com/taleinat/fuzzysearch) so that it would match close matches instead of just the exact words that were in the search.
Or to get even more ability, it could use one of the free neural network models that create vectors from text (for both the text of books and for your search term) and match on the nearest vectors. This would be able to match related ideas even if they don't use the same words.
I know this seems like a pi in the sky idea, but I thought I would put it out here anyway. Here's hoping!
Thanks for this amazing software! Ok