Is your feature request related to a problem? Please describe.
Semantic chunking cuts into sentences and then merges those with close semantic scores (cosine similarity) to form passages.
However, when parsing on a page-by-page basis, semantic score comparisons are only made between sentences on a page, which may not perform as well as comparing sentences across the document.

Describe the solution you'd like
Implement a feature to parse on a per-document basis when doing semantic chunking, even if you're using a per-page parsing module (e.g. clova, table_hybrid_parse)