Issue with Periods in Dates or numbers Causing Incorrect Segment Splitting in German Transcriptions

[sijitang]

When transcribing German audio files with WhisperX and using whisperx.load_align_model, I noticed that dates containing periods (e.g., "Mai.2022") or number containing periods are incorrectly split into separate segments. This occurs because the period is interpreted as the end of a sentence, which leads to inaccurate time alignment and text segmentation.

00:00:59,879 --> 00:01:00,039] Die 5.
[00:01:00,479 --> 00:01:12,850] Strafkammer sah es als erwiesen an, dass der 52-Jährige wissentlich eine verbotene Nazi-Parole bei einer AfD-Veranstaltung im Mai 2021 verbreitet hatte.

[00:14:13,689 --> 00:14:15,992] Und nun die Wettervorhersage für morgen Mittwoch, den 15.
[00:14:16,032 --> 00:14:16,093] Mai.