All projects

Case study 04 / University project · Applied NLP

Healthcare NLP System

Clinical language, made inspectable.

An 8-tab Streamlit application for clinical entity recognition, relation extraction and summarisation.

  • BioBERT
  • ClinicalBERT
  • Streamlit
  • NLP
  • Fine-tuning

01 / Context

Problem

Clinical text contains entities, relationships and narrative information that need different processing tasks. This project brings named-entity recognition, relation extraction and clinical summarisation into one application.

The work combines fine-tuned BioBERT and ClinicalBERT with an 8-tab Streamlit interface.

02 / Ownership

My Role

Built the healthcare NLP application, bringing the supplied model fine-tuning and language-processing tasks into an 8-tab Streamlit interface.

The public repository is available for inspection. Exact evaluation scores, datasets and individual/team contribution boundaries should be checked against its documentation.

03 / System design

Architecture

  1. 01Clinical text

    Input to the NLP application

  2. 02Model layer

    Fine-tuned BioBERT / ClinicalBERT

  3. 03Language tasks

    NER · Relation extraction · Clinical summarisation

  4. 04Application

    8-tab Streamlit interface

Conceptual view of the supplied components. It does not assign a specific model to an unconfirmed task or reproduce real patient records.

04 / Engineering decisions

Engineering Decisions

Showing the work beyond the model.

The project combines model work with an application: entity recognition, relation extraction and summarisation are exposed through the Streamlit interface.

The repository is the inspectable artefact. No clinical validation or accuracy claim is made without the evaluation protocol and results.

05 / Evidence

Results

8application tabs
NERentity recognition
Codepublic repository available

The demonstrated result is the application and its supplied capabilities. Benchmark scores and clinical-use validation have not been supplied.

06 / Tools & capabilities

Stack

  • BioBERT
  • ClinicalBERT
  • Streamlit
  • NLP
  • Fine-tuning
Next case study

ROS2 Robot Vision & Navigation