Zero-Shot Entity Extraction

U

@

·

Zero-shot entity extraction using only type definitions with span identification, confidence scoring, and disambiguation handling.

24 copies0 forks
Extract entities from text without training examples, using only entity type definitions.

## Text to Process
{{text}}

## Entity Types
{{entity_definitions}}

## Extraction Rules
{{extraction_rules}}

**Extraction Instructions:**

1. Read the text carefully, identifying spans that match entity definitions.

2. For each potential entity:
   - Verify it matches the type definition
   - Extract the exact text span
   - Identify start/end positions
   - Assign confidence level

3. Handle edge cases:
   - Overlapping entities
   - Ambiguous types
   - Partial matches

**Output Format:**
```json
{
  "entities": [
    {
      "text": "extracted text",
      "type": "entity_type",
      "start": 0,
      "end": 10,
      "confidence": 0.95
    }
  ]
}
```

Include disambiguation notes for ambiguous cases.

Details

Category

Coding

Use Cases

Entity extractionNERText processing

Works Best With

claude-sonnet-4-20250514gpt-4o
Created Shared

Create your own prompt vault and start sharing