Finding a chunking method for PDFs with different structures
A RAG for university and school documents needs a method that works across very different PDF structures. Some PDFs have useful headings, while others do not.
Automatically detected titles can also be wrong, which weakens the search results. No tested solution, comparison, or numbers are provided.
Key points
- The is intended for university and school documents.
- PDFs differ in whether they contain re headings.
- Incorrectly detected titles can reduce search quality.
- No tested method or data is included.