AI-powered tools for hospital clinical informatics — from NLP-driven chatbots to patient data insights, supporting clinicians with context-aware, evidence-based information systems.
Available Tools
Research
Machine Learning Chatbot Development: Bidirectional LSTMs, Advanced NLP Techniques, and Context-Aware Design in a Clinical Information System
This paper details the development of the Hypotify Clinical Insights Bot, a context-aware chatbot built with a Bidirectional LSTM architecture, VADER sentiment analysis, spaCy NER, and a 3-turn dialogue state tracker. Trained on 370 labeled clinical intent examples across 20 categories, the system integrates with real patient demographics, hospital admissions, ICD-10 diagnoses, and lab results to deliver live clinical data insights.
Chatbot Model Development and Training: Bidirectional LSTMs, Word2Vec Embeddings, and Context-Aware Response Design
This paper walks through the construction of a machine learning chatbot for clinical data insights. It covers the training dataset (370+ labeled examples across 20 intent categories), a 9-step preprocessing pipeline with lemmatization and TF-IDF analysis, Word2Vec embeddings, and a Bidirectional LSTM architecture chosen for its ability to capture contextual dependencies in both directions. The model achieved 97.3% validation accuracy with a context-aware response system integrating VADER sentiment analysis, spaCy NER, and a 3-turn dialogue state tracker over real patient demographics, hospital admissions, ICD-10 diagnoses, and lab results.
Databases Meet Conversational AI: SQLite Integration, Dynamic Knowledge Retrieval, and the Architecture of a Clinically-Aware Data-Driven Chatbot
This paper documents the database integration of the Hypotify Clinical Chatbot — redesigning the core data layer from CSV-based file reads to a structured SQLite database. It covers importing 100,000 patient records, 361,760 admissions, 361,760 diagnoses, and 100,000 lab results into an indexed relational schema with foreign key enforcement. The architecture implements the Repository pattern through db_manager.py, conversation logging for auditability, a user feedback mechanism, and a fallback hierarchy ensuring graceful degradation. Query performance improved from 2–4 seconds (full CSV reads) to under 5 milliseconds (indexed SQLite lookups).
Accuracy Auditing, Behavioral Fixes, and What It Means to Test a Chatbot Honestly
This paper documents the post-Week 10 accuracy audit of the Hypotify Clinical Chatbot. After running 100 questions against the live Streamlit engine, the initial pass rate was 86%. The 14 failures fell into three categories: non-deterministic responses (7), wrong intent routing (5), and inaccurate expected answers (2). Root cause analysis traced each failure through the chatbot's _build_response() method. Five targeted code fixes — explicit help/farewell intercepts, non-hex UUID handling, expanded population insights vocabulary, and NER capability routing — achieved a 100% final pass rate with 20 documentation corrections.