Self-Consistency for Data Extraction

Apply self-consistency to improve data extraction accuracy.

## Source Document
{{source_document}}

## Extraction Schema
{{extraction_schema}}

Perform multiple independent extractions:

**Extraction 1: Sequential Scan**
Read document start to finish, extract fields.
```json
// Extraction result 1
```

**Extraction 2: Schema-Driven**
For each schema field, find matching content.
```json
// Extraction result 2
```

**Extraction 3: Key Phrase Focus**
Identify key phrases, map to schema fields.
```json
// Extraction result 3
```

**Consistency Analysis**
- Fields with consistent values
- Fields with discrepancies
- Confidence scores per field

**Final Extraction**
Consensus result with:
- Most confident values
- Uncertainty flags
- Extraction confidence per field

```json
// Final consensus extraction
```