LLM结构化信息提取与实体标注模板

You are an expert information extraction agent. Your task is to extract structured entities from the following unstructured text.

Extraction Schema

For each entity found, output:

• entity_class: The category (e.g., person, organization, date, metric, relationship)
• entity_text: The exact verbatim text from the source (do NOT paraphrase)
• attributes: A JSON object with relevant attributes for context
• source_location: Approximate position in the text for traceability

Rules

1. Use EXACT text spans from the source document — never paraphrase or summarize
2. Do not create overlapping entity spans
3. Process the document in order of appearance
4. If an entity appears multiple times, extract each occurrence separately
5. Provide meaningful attributes that add context beyond the raw text

Few-Shot Example

Input: "Apple Inc. reported Q3 revenue of $81.8 billion, a 5% increase year-over-year, announced CEO Tim Cook."

Output:

[
  {"entity_class": "organization", "entity_text": "Apple Inc.", "attributes": {"type": "public_company", "sector": "technology"}},
  {"entity_class": "metric", "entity_text": "$81.8 billion", "attributes": {"metric_type": "revenue", "period": "Q3", "change": "+5% YoY"}},
  {"entity_class": "person", "entity_text": "Tim Cook", "attributes": {"role": "CEO", "organization": "Apple Inc."}}
]

Your Task

Now extract all entities from the following text. Output valid JSON array only.

[PASTE YOUR TEXT HERE]