File size: 6,663 Bytes
fcaa164 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 |
system_prompt: |
You are a helpful academic expert and web developer, who is specialized in generating a paper project page, from given research paper's contents and figures.
template: |
Below is the raw content with markdown text, images, and tables information:
<raw_content>
{{raw_content}}
</raw_content>
Your task is to analyze the paper content and generate a complete structured full_content JSON that contains ALL the information needed for the final HTML webpage. This JSON will be the single source of truth for generating the project page.
You need to:
1. **Extract Paper Metadata**:
- Paper title
- Authors with their affiliations (use <sup> tags for superscripts)
- Institution affiliations
- Any other relevant metadata (conference, year, links, etc.)
2. **Analyze and Plan Paper Sections**:
- Identify the main sections from the paper (Abstract, Introduction, Method, Results, Analysis, Conclusion, etc.)
- For each section, extract the key content that should appear on the project page
- Write clear, concise content summaries that will be displayed
- DO NOT just copy raw paper text - synthesize and adapt it for web presentation
3. **Select and Place Visual Elements**:
- Identify the teaser figure (the most important visualization, usually appears first)
- For each section, select the most relevant figures and tables
- Use the EXACT paths provided in raw_content for all images and tables
- Include the exact width and height values from raw_content
- Write descriptive captions for each visual element
- Each figure/table should be used at most once
- Ensure all important figures are included
- For sections with multiple tables, choose only the most relevant one
4. **Content Guidelines**:
- The teaser figure must be included and should appear early (typically in Overview or after Abstract)
- Prioritize pictures and tables based on their relevance and importance
- Ensure figures are closely related to their section's content
- Match visual elements with their corresponding text discussions
- Specify clear placement instructions for each visual element
- Write content that flows naturally and is appropriate for a web page (not raw academic text)
5. **Path and Dimension Requirements**:
- Use EXACTLY the same paths as provided in raw_content (e.g., "assets/paper-picture-8.png")
- Include the exact width and height values from raw_content
- Maintain the original aspect ratios of all visual elements
Please provide your complete full_content structure in the following JSON format:
```json
{
"title": "Complete paper title",
"authors": "Author names with <sup> tags for affiliations, e.g., 'John Doe<sup>1</sup>, Jane Smith<sup>2</sup>*'",
"affiliation": "Complete affiliation text with <sup> tags, e.g., '<sup>1</sup>MIT, <sup>2</sup>Stanford University'",
"teaser_figure": {
"path": "exact path from raw_content",
"description": "detailed description of the teaser figure",
"width": "width value from raw_content",
"height": "height value from raw_content",
"caption": "caption text for the teaser"
},
"Section Name 1": "Complete content text for this section. This should be well-written, web-appropriate content that synthesizes the paper's key points. Include inline references to figures like: [Figure description][path][width=X, height=Y](figure_number) when you want to reference a visual element.\n\n![Detailed caption describing what the figure shows][assets/exact-path.png][width=1234, height=567](1)",
"Section Name 2": "Content for the next section with its own flow and structure...\n\n![Another figure caption][assets/another-path.png][width=890, height=456](2)",
"Section Name 3": "More content...\n\n![Table caption][assets/table-path.png][width=2000, height=800](3)"
}
```
CRITICAL Requirements for the JSON structure:
1. **Metadata Fields** (required at the top):
- "title": The full paper title
- "authors": Author names with superscript affiliations
- "affiliation": Institution information with superscripts
- "teaser_figure": A separate object with path, description, width, height, and caption
2. **Section Fields** (one per major paper section):
- Use clear section names as keys (e.g., "Overview", "Method", "Experimental Results")
- Each section's value should be a string containing:
* Well-written, web-appropriate content that explains the section
* Embedded figure/table references using the notation: ![caption][path][width=X, height=Y](number)
* The figure notation MUST be on a new line (with \n\n before it)
* Natural flow and transitions between content and figures
3. **Figure/Table Notation Format**:
- Use: ![Caption text][exact/path/from/raw_content][width=1234, height=567](figure_number)
- The figure_number must be a unique integer (1, 2, 3, ...)
- Caption should describe what the visual shows
- Path must EXACTLY match raw_content
- Width and height must EXACTLY match raw_content
- Place figures after the relevant text that discusses them
4. **Content Writing Guidelines**:
- Write clear, engaging content suitable for a project page (not raw academic prose)
- Each section should tell a coherent story
- Ensure smooth transitions between text and visuals
- Highlight key contributions and findings
- Keep the tone professional but accessible
- DO NOT just copy-paste from the paper - adapt and synthesize
5. **Visual Placement Strategy**:
- Teaser figure: Separate field, will be placed prominently at the top
- Section figures: Embedded in section text where most relevant
- Place figures after the text that introduces or discusses them
- Ensure balanced distribution of visuals across sections
- Don't overload any single section with too many visuals
Important reminders:
- All paths must EXACTLY match those in raw_content
- All width and height values must EXACTLY match those in raw_content
- Figure numbers should be sequential and unique across the entire document
- Each visual element should appear only once
- The teaser figure should be the most impactful/representative visualization
- Section names should be clear and match the paper's structure
- Content should be web-friendly, not just copied academic text
- Use \n\n before figure notations to ensure they're on new lines in the JSON string
jinja_args:
- raw_content |