Extracting Medical Attributes from Clinical Trials

on YouTube ↗26m

" summary

Extracts diseases, dosages, ages, genders and side effects from FDA drug labels and clinical trials using noun/phrase candidates, word2vec/Gensim, the Stanford NLP parser and simple classifiers; a curated seed of ~8,000 drug/disease names plus hand-labeling ~100 candidates lets rule-based bootstrapping produce a reduced dataset with ≈95% accuracy, then a logistic regression model yields high precision/recall for drug→disease links. Integrates weak supervision via DeepDive and IPython notebooks; pipeline is deployable (Spark-ready) and the UI supports rapid correction and iteration.

AI-written summary · about

" chapters (8)

1. 0:00Introduction and motivation
2. 2:21FDA drug labels analysis
3. 6:13Training data challenges
4. 10:04Machine learning pipeline
5. 13:55Clinical trials extraction
6. 18:14DeepDive tool overview
7. 22:59Conclusion and future work
8. 24:02Q&A and lowest hanging fruit

" related talks

• 47mAI for the Proper Use of MedicationsSuzanne Bento Pereira, Romain Benassi · Devoxx France 2022
• 20mDiagnostic Medical application of MLCarrie Weinberg · MLconf NYC 2019
• 26mNatural Language Processing for Healthcare: Current Trends...Amir Tahmasebi · Global Artificial Intelligence Conference 2020
• 37mCreating Knowledgebases from unstructured textSanghamitra Deb · PyData San Francisco 2016
• 30mLeveraging Molecular Data for Drug DiscoveryHaiyan Huang · Computational Challenges in Very Large-Scale 'Omics'
• 32mFail Fast, Save Millions: AI for Smarter Clinical TrialsPiotr Ćwikliński · Confitura 2025
• 36mFormatting Biological Big Data to Enable (Personalized) Systems PharmacologyPatrick Aloy · From Algorithms to Discovery in Genome-Scale Biology and Medicine 2022
• 33mVisualizing Machine Learning of Units of Measure using PyVizJacob Barhak · PyData Austin 2019