Finding relations in documents with IEPY
summary
IEPY is an open source Python system that combines a rule-based + machine learning hybrid pipeline with a document store, a learning engine, a web UI and scikit-learn SVM classifiers. It extracts binary relations from text, successfully pulling funding events from TechCrunch funding extraction and doubling detected entities when classifiers augment rules, and it helps link people in military files for desaparecidos. It works best on medium, fairly regular corpora, struggles with memory-heavy training, larger arity relations and event temporal disambiguation.
AI-written summary · about
chapters (7)
- 0:00Introduction
- 10:36IEPY overview and architecture
- 15:17Manual and rule-based extraction
- 17:55Machine learning for relations
- 21:11Argentine military files case
- 24:08IEPY strengths and limits
- 29:35Summary
related talks
25mInventing Curriculum using Python and spaCy
23mPython for Threat Intelligence
35mUsing Scattertext and the Python NLP Ecosystem for Text Visualization
28mIbis framework, Making data science work at any scale
36mPython and ETL Schemas: The essential pivot for Artificial Intelligence
05mHow to build a Code Search Engine
39mKeynote 3: How did you know? Explaining Black Box Model Predictions in Python
37mSpark Backend for Ibis: Seamless Transition Between Pandas...