Back to Discover
🚀 Extract Headings from Document
Extract Headings from Document description placeholder
Prompt
(Prompt 7) Please analyze the following text from a document and extract ONLY the true section headers. Structure the output in a hierarchical markdown format, using # for H1, ## for H2, ### for H3, etc.
A true header in this document typically:
1. Is short (usually 1-7 words)
2. Does not form part of a complete sentence
3. Introduces a new section of content
4. Often appears in bold, ALL CAPS, as 1-5 words on a line with blank lines on either side, or larger font than surrounding text
5. Is sometimes in bold and followed by a colon and then the regular text of a paragraph (e.g., "**Header/Title:** This part is just paragraph text.")
DO NOT include as headers:
- The first few words of a paragraph even if they appear in bold
- Sentence fragments that are part of longer sentences or paragraphs
- Individual bullet points
- Individual numbered/alpha-numeric bullet points (e.g., "a. IEP changes should be made" or "1. Placement changes should happen.")
- Phrases that would not make sense as standalone section titles
- Your own summary of what you think a section of text is about
For headers that appear as "**Title:** Content", extract ONLY the "Title" part as the header.
Determine the appropriate hierarchy level based on:
- Font size/styling in the original document (if detectable) (e.g., centered headers would be higher in hierarchy than left-aligned headers)
- Semantic relationships between headers
Output ONLY the headers in a clean hierarchical markdown structure with no explanatory text.
DO NOT CREATE HEADERS that you think should or could exist. Only use the text that is in the document as a header.
<document>
{{ document }}
</document>