Hypotify Clinical Insights Bot
Hypotify is a machine learning chatbot built with a Bidirectional LSTM architecture for intent classification across 20 clinical categories. It integrates sentiment analysis, named entity recognition, and a 3-turn context window to deliver context-aware responses over real patient demographics, hospital admissions, ICD-10 diagnoses, and lab results.
Tech Stack
Interactive web interface for chatbot interaction
Core backend, ML pipeline, and chatbot logic
Patient data loading, querying, and live statistical analysis
Numerical operations for model input preparation
Bidirectional LSTM model for intent classification
Tokenization, lemmatization (WordNet), stop-word removal, and VADER sentiment analysis
Named entity recognition using en_core_web_sm pipeline
TF-IDF vectorization, train/test splitting, and label encoding
Word embeddings trained on clinical corpus (64-dim, CBOW)
Key Features
Bidirectional LSTM intent classification across 20 clinical categories
VADER sentiment analysis on every user message
spaCy named entity recognition (patient IDs, dates, organizations)
3-turn dialogue state tracking for context-aware follow-ups
Live patient data queries (100K patients, 361K admissions, 1M+ lab results)
Confidence-based fallback with 0.45 threshold for clinical safety
ICD-10 diagnosis lookup and top-10 frequency analysis