[BUG] Docling HTML to Markdown conversion loses structure on members page

[niloydebbarma-code]

Is there an existing issue for this?

• I have searched the existing issues

Current Behavior

Problem

After running 1_crawl_site --url https://thealliance.ai/ then 2_process_files, company names get merged into one long line. Can't tell where one name ends and another starts when checking the processed markdown.

How to reproduce

1. Run 1_crawl_site --url https://thealliance.ai/
2. Run 2_process_files
3. Check workspace/processed/thealliance.ai__aia-members.md
4. Find that all organization names are concatenated together

What happens

HTML has each company separated:

<span>AMD</span>
<span>IBM</span>  
<span>Meta</span>
<span>Oracle</span>

After docling processes it, they run together:

AMD IBM Meta Oracle Sony Uber Dell Technologies ServiceNow Snowflake Databricks Cornell University...

Impact on entity extraction

These get extracted:

• Acceleration Consortium (starts the line)
• Carnegie Mellon University (has "University")
• Meta (short and clear)
• Red Hat (distinctive)

These get lost:

• AMD (buried in "...AI AMD Anaconda...")
• IBM (buried in "...Jerusalem IBM Imperial...")
• Intel (buried in "...Technology Intel The...")
• Sony (buried in "...Snowflake Sony St Johns...")

Files affected

• Source HTML: workspace/crawled/thealliance.ai__aia-members.html
• Processed markdown: workspace/processed/thealliance.ai__aia-members.md
• Processing script: 2_process_files.py (uses docling conversion)

Folder structure

workspace/
├── crawled/           # Raw HTML files from website crawl
└── processed/         # Docling-converted markdown files

Fix options

1. Change docling settings to preserve HTML structure
2. Parse HTML pages separately before docling conversion
3. Post-process concatenated text to split entity names properly

Expected Behavior

No response

Steps To Reproduce

No response

Environment

- OS:
- Browser:

Anything else?

No response