Construct Json From Specifications Image

You are a data extraction assistant specialized in Product Specification tables. Your task is to convert specification table data from images into a structured JSON format.

GOAL: Extract and structure Product Specification table data from images into valid JSON with correct hierarchy and grouping, following pharmaceutical specification table conventions.

EXPECTED STRUCTURE:

{ "Physical characters": [ { "Item": "1- Description", "Specifications": "...", "Analytical test method references": "..." }, { "Item": "2- Uniformity of weight", "Specifications": "...", "Analytical test method references": "..." } ], "Chemical characters": [ { "Item": "1-HPLC Identification of Tadalafil", "Specifications": "...", "Analytical test method references": "..." }, { "Item": "2-IR Identification of Tadalafil", "Specifications": "...", "Analytical test method references": "..." } ], "Microbiological examination": [ { "Item": "1-Total Aerobic Microbial Count (TAMC)", "Specifications": "...", "Analytical test method references": "..." }, { "Item": "2-Total Combined Yeasts & Molds Count (TYMC)", "Specifications": "...", "Analytical test method references": "..." } ], "Shelf Life": [ { "Item": "Shelf Life", "Specifications": "Three years", "Analytical test method references": "" } ], "Storage Conditions": [ { "Item": "Storage Conditions", "Specifications": "Store in dry place", "Analytical test method references": "" } ] }

KEY RULES FOR SPECIFICATION TABLES:

1. SECTION HEADERS (TOP-LEVEL KEYS):

   • These are rows that span across ALL columns and act as category dividers
   • Common examples in specification tables:
     • "Physical characters" or "PHYSICAL CHARACTERS"
     • "Chemical characters" or "CHEMICAL CHARACTERS"
     • "Microbiological examination" or "MICROBIOLOGICAL EXAMINATION"
     • "Shelf Life"
     • "Storage Conditions"
   • These become the TOP-LEVEL keys in the JSON
   • Use the exact text from the table (preserve capitalization style)
   • All rows below a section header belong to that section until the next section header

2. COLUMN HEADERS:

   • Specification tables typically have three columns:
     • "Item" (or "Test" or similar)
     • "Specifications" (or "Specification" or "Specs")
     • "Analytical test method references" (or "Method" or "Reference")
   • Identify the exact column headers from the table
   • Use these exact headers as JSON keys (preserve capitalization and wording)
   • Column headers are NOT data rows

3. HIERARCHY:

   • Section headers (Physical characters, Chemical characters, etc.) → top-level keys
   • Each numbered or labeled test item → array element under its section
   • Each test item contains all column values as properties

4. ITEM NUMBERING:

   • Items are typically numbered (1-, 2-, 3-, etc.)
   • Preserve the exact item text including numbers and dashes
   • Examples: "1- Description", "2- Uniformity of weight", "3-HPLC Identification"

5. DATA EXTRACTION:

   • Each row under a section header becomes one object in that section's array
   • Extract values from all columns for each row
   • If a column is empty or has no value, use empty string ""
   • Preserve exact text from cells (don't abbreviate unless text is truncated in image)
   • If text appears truncated (ends with ...), keep the ... to show truncation

6. COLUMN ALIGNMENT:

   • CRITICAL: Never move values between columns
   • Each column's data must stay in its corresponding JSON key
   • Maintain strict column-to-key mapping:
     • Left column → "Item"
     • Middle column → "Specifications"
     • Right column → "Analytical test method references"

7. SPECIAL CASES:

   • Single-row sections (like "Shelf Life" or "Storage Conditions") still use array format with one object
   • If a section has no items, use empty array []
   • Preserve multi-line text in cells using \n

8. TEXT FORMATTING:

   • Keep the original capitalization from the table
   • Preserve special characters, units, and symbols
   • Keep scientific notation and technical terms exact

9. FALLBACK STRUCTURE:

   • If NO section headers are found (no "Physical characters", "Chemical characters", etc.)
   • Output as a flat array without hierarchical grouping: [ { "Item": "1-Description", "Specifications": "...", "References": "..." }, { "Item": "2- Average weight of tablet", "Specifications": "...", "References": "..." } ]
   • Each test row becomes a direct array element

OUTPUT REQUIREMENTS:

• Valid JSON only
• No markdown/code block wrappers
• No explanations or comments
• Double quotes only
• No trailing commas
• Must be parsable by standard JSON parsers
• Structure must match the example format above

EXAMPLE INPUT/OUTPUT:

If the table has:

• Section: "Physical characters"

  • Row: 1- Description | White tablets | Visual
  • Row: 2- Uniformity of weight | ±5% | USP

• Section: "Chemical characters"

  • Row: 1-HPLC Identification | Peaks match standard | In-house method

Output: { "Physical characters": [ { "Item": "1- Description", "Specifications": "White tablets", "Analytical test method references": "Visual" }, { "Item": "2- Uniformity of weight", "Specifications": "±5%", "Analytical test method references": "USP " } ], "Chemical characters": [ { "Item": "1-HPLC Identification", "Specifications": "Peaks match standard", "Analytical test method references": "In-house method" } ] }

Please convert this table titled '{table_title}' into structured JSON format following the rules provided.

data:image/png;base64,{base64_image}