docling_parse_v2 split/connect words

[InbarShapira]

Bug

There are cases that docling_parser_v2 spilt words to it characters or connect words

Example1:
Original text: products that were recently iroduced
markdown: products that were re c e n t l y i roduced

Example2:
Original text: Tables 2–5 show the results of partitioning the graphs in our test suite on
markdown: Tables 2-5 sho w theresultsfpartitioningegraphsinourtest suite on

Steps to reproduce

...

Docling version

Docling version: 2.21.0
Docling Core version: 2.18.0
Docling IBM Models version: 3.3.0
Docling Parse version: 3.3.0
Python: cpython-311 (3.11.4)
Platform: macOS-14.6.1-arm64-arm-64bit

Python version

Python 3.11.4

[PeterStaar-IBM]

@InbarShapira Can you please provide an example?

[dhdaines]

This is almost certainly docling-project/docling-parse#99

[dhdaines]

If you downgrade to docling-parse 3.1.2 (you will also need to downgrade docling to 2.21 at least I think) does the issue still occur?

[dhdaines]

you can also of course try this PR! docling-project/docling-parse#105