PeterPaker123/Qwen2.5-7B-Vietnamese-Medical-NER-GRPO

Overview

This model, developed by PeterPaker123, is a Qwen2.5-7B based generative Large Language Model (LLM) specifically fine-tuned for Vietnamese Medical Named Entity Recognition (NER). Unlike traditional NER models with fixed entity sets, this model functions as an intelligent, prompt-driven extraction agent. It processes Vietnamese medical texts and outputs highly structured JSON lists of extracted entities.

Key Capabilities

• Dynamic Entity Extraction: The model can identify and extract any medical, clinical, or demographic entities based on user-defined instructions in the system prompt, rather than being limited to a predefined set of labels.
• Context Flagging: Beyond simple entity extraction, it can assign assertion attributes (context flags) such as is_negated, is_uncertain, severity, temporality, is_historical, is_family, and is_conditional to each extracted entity, providing richer contextual information.
• Structured Output: All extractions are provided in a clean, parseable JSON format, making integration into downstream applications straightforward.
• Vietnamese Language Focus: Exclusively trained and optimized for the Vietnamese language, leveraging a consolidated corpus from four premier Vietnamese medical datasets (VietBioNER, PhoNER_COVID19, ViMQ, ViMedNER).

What Makes This Model Different?

This model distinguishes itself by transitioning from traditional token-classification BERT models to an LLM-based approach for NER. Its core innovation lies in its dynamic extraction capability, achieved through Supervised Fine-Tuning (SFT) where target entity schemas were injected directly into system prompts. This method taught the model to generalize the skill of extracting exactly what the prompt asks for, rather than merely memorizing a hardcoded list of labels. This allows for highly flexible and adaptable medical NER in Vietnamese, supporting a wide range of user-defined entity types and contextual attributes.

Use Cases

This model is ideal for applications requiring precise and flexible medical information extraction from Vietnamese text, including:

• Enriching medical chatbot prompts in Retrieval-Augmented Generation (RAG) pipelines.
• Automating the structuring of telehealth transcripts and clinical notes.
• Constructing clinical knowledge graphs.
• Supporting epidemiological data mining and surveillance efforts.

Note: This model is an AI research tool and must not be used for automated medical diagnosis, treatment planning, or life-or-death triage without human oversight.