Machine Learning Internship

Project Overview

Developed and implemented Natural Language Processing (NLP) models to analyze medical consultation texts, specifically focused on detecting inappropriate behavior or instances of violence in gynecologist-patient interactions. This project aimed to improve patient safety and healthcare quality by identifying concerning patterns in medical communications.

BERT Architecture

Text Classification Results

Text Preprocessing Pipeline

Model Performance Comparison

Key Responsibilities

• Selected and experimented with various NLP model architectures to determine the most effective approach for medical text analysis, with a focus on transformer-based models.
• Collected and preprocessed a labeled dataset of medical consultation texts, including data cleaning, tokenization, and normalization to prepare for model training.
• Developed specialized algorithms to detect instances of violence or inappropriate behavior in gynecologist-patient interactions through contextual language understanding.
• Fine-tuned BERT models on medical domain texts to improve detection accuracy and sensitivity to subtle linguistic indicators of problematic interactions.
• Implemented evaluation metrics specific to healthcare applications to ensure model reliability and minimize false positives/negatives in critical patient safety contexts.