hi i am trying to extract the information from the pdf but my pdf tables and columns and hierarchy tables can help me on this how to do

[harshithgowdakc]

hi i am trying to extract the information from the pdf but my pdf tables and columns and hierarchy tables can help me on this how to do

[JorjMcKie]

Sorry, I do not understand at all what you mean. Please reword your post.

[Harshith9898]

I'm trying to extract information from a PDF that contains tables, columns, and hierarchical structures. I'm having trouble preserving the layout and structure during extraction.

[JorjMcKie]

Please be specific and attach one (small) of your problem files.
Also explain what you mean by "preserving layout & structure" across all those different object types.

[Harshith9898]

When I try to extract text from the PDF, I get output like this:

"Software Specifications
Operating System
3rd-Party Integration
Server-Side Database
Suggested Browser for Client-Side
Windows7/8/10/11
Notification / Messages
Line / WhatsApp / Amazon SNS / SMS"

But I want it to be extracted like this:

"3rd-Party Integration Notification / Messages Line / WhatsApp / Amazon SNS / SMS"

Let me know if you'd like help figuring out how to achieve this clean extraction too. i want to do chunk on next step for LLM and RAG

[JorjMcKie]

I am a bit out of ideas how to say this in order to reach you:
Please provide the PDF example and the code you tried!
A picture alone does not help at all.

[JorjMcKie]

Sorry, now that you shared your code, I realize that you are not using PyMuPDF or PyMuPDF4LLM at all, but other packages.
You are posting in the wrong repository.
Please submit a problem at pypdf's repository.

[harshithgowdakc]

i am using PyMuPDF also import pymupdf4llm
from langchain.text_splitter import MarkdownTextSplitter

Get the MD text

md_text = pymupdf4llm.to_markdown("/content/ZKBio CVSecurity_V6.4.0_R_Datasheet_202411.pdf") # get markdown for all pages

splitter = MarkdownTextSplitter(chunk_size=1200, chunk_overlap=200)

splitter.create_documents([md_text])
ZKBio CVSecurity_V6.4.0_R_Datasheet_202411.pdf
the output i am getting like this Document(metadata={}, page_content="-----\n\n|Col1|Col2|Col3|\n|---|---|---|\n|Windows7/8/10/11, Windows Server 2008/2012/2016/2019/20|22|Notification / Messages|\n|||GIS Map|\n|||Microsoft Active Directory|\n|||Intrusion Alarm Integration|\n|PostgreSQL, Oracle11g/1 19c /21c, SQL Server 2008 014/2016/2017/2019/202|2c/18c/ /2012/2 2|API|\n|IE 11 or later, Chrome 33 or later, Safari 6.1.3 or later Edge|||\n|ISO/IEC 27001:2013, ISO/IEC 27701:2019, ISO9001, ISO20000.|||\n|Unlimited (depending on the performance of the server & network)|||\n|300,000|||\n|300,000|||\n|300,000|||\n|300,000|||\n|300,000|||\n|5,000|||\n|Unlimited (depending on the performance of the server & network)|||\n|2,000|||\n|100,000|||\n|1024|||\n|50||Elevator Destination Control System Integration|\n|Unlimited (depending on the device's configuration)|||\n|||Data Protection|\n\n\n\n\n\n\n\n-----"),

[harshithgowdakc]

help me

[JorjMcKie]

Sorry, there is no way that PyMuPDF4LLM can handle this page correctly. The table is extremely complex plus the page contains background / watermark images which together simply make this impossible.
You will have to develop your own logic.

[JorjMcKie]

Try to use conventional text extraction, maybe in some layout preserving way like Page.get_text(sort=True).

[Harshith9898]

ok Thnks for that