Dharma Field Extractor: Structured JSON Output from Any Document
Story 1, from Hugging Face
Dharma Field Extractor is a 2-billion-parameter model that extracts structured fields from documents into JSON, outperforming general-purpose LLMs in quality and cost.
Key points
- Extracts structured fields from documents into typed JSON
- Outperforms general-purpose LLMs on field extraction tasks
- Supports both open-ended and schema-directed extraction
- Trained on multi-domain data including Brazilian and multilingual documents
- Runs on a single consumer GPU at lower cost than frontier APIs
Why it matters
Software engineers dealing with document processing can benefit from structured field extraction, reducing manual data mapping and improving integration with downstream systems.
Counterpoint
The model requires training data with field-level annotations, which may not be available for all document types.
Sources
This summary was written by a language model and may contain errors. Read the original for the full story.