How Machine Learning and AI Can Support the Fight Against COVID-19
summary
Uses the CORD corpus of ~400,000 papers (≈200,000 full text) and processes documents in 500‑paper chunks, parallelizing via Azure ML sweep jobs to avoid 11–12 hour single‑node runs; trains a PubMed‑tuned BERT for token classification on BC5‑CDR (COVID tokens missing because training data is older), and also runs Text Analytics for Health for entity/relation extraction; stores ~400k JSON records in Cosmos DB and visualizes time trends (hydroxychloroquine spike then drop) with Power BI — pipeline works but domain‑specific NER is needed for deeper studies.
AI-written summary · about
chapters (14)
- 0:00Start
- 3:52CORD Papers Dataset
- 6:35How BERT Works (Simplified)
- 8:15Main Idea
- 9:12Part 1: Extracting Entities and Relations
- 14:16Azure Machine Learning
- 28:06Entity Extraction
- 29:14Using Text Analytics for Health
- 31:41Analyzing CORD Abstracts
- 33:00Parallel Sweep Job in Azure ML
- 36:19Cosmos DB & Azure Data Solutions
- 37:27Cosmos DB SQL Queries
- 44:00Further Exploration Jupyter in Cosmos DB
- 44:35How Medication Strategies Change
related talks
37mUsing the power of Azure and AI to help fight the COVID pandemic
28mNLP Solutions for infodemic management
22mCOVID-19: Data Science & Expertise
29mAI for COVID-19: An online virtual care approach
55mTake Control of the Data of You
26mAccelerating COVID-19 insights for clinical use with intelligent search
37mAI/ML Strategies & Techniques for a Post Pandemic World
23mLeveraging Generative AI and Transformers in Healthcare