Closed CypherousSkies closed 2 years ago
Nope. Well kinda. The problem is TTS.utils.synthesizer.Synthesizer.split_into_sentences only really working for english sentences for some reason. In the french text I'm testing it says that a whole paragraph (4k+ characters) is a single sentence. It's not. So time to override with nltk. Hopefully. On a day that's not today.
ok update it's not TTS's split into sentences. it likely has something to do with malformed headers and footers, as well as excessive spaces in the filtered text. Would likely be fixed by #9
Closing for being redundant with #9 (and so that I feel better about releasing 0.0.4)
Scans of french language texts have non-linguistic artifacts (like how english texts have ^L) Just gotta find them and add them to TextProcessor's filter