bug: different pdf text extraction output from `unstructured` and `unstructed-inference` libarary

[christinestraub]

Describe the bug PDF text extraction by pdfminer works differently in unstructured repo and unstructured-inference repo.

To Reproduce PDF: algebra-graph-level1-1.pdf

• unstructured-inference repo

  
  # unstructured-inference repo

layout = process_file_with_model( filename="algebra-graph-level1-1.pdf", model_name=None, )

elements = layout.pages[0].elements

print("\n\n".join([str(el) for el in elements])) print(f"n_of_elements: {len(elements)}")

- unstructured repo

unstructured repo

elements = partition_pdf( filename="algebra-graph-level1-1.pdf", strategy="fast", )

print("\n\n".join([str(el) for el in elements])) print(f"n_of_elements: {len(elements)}")

**Expected behavior**
The PDF text extraction output should be the same.

**Additional context**
The logic for extracting PDF text using `pdfminer` was implemented separately in both repositories.

[christinestraub]

Completed via #2158