Explore strategies for successful Natural Language Processing (NLP) projects in this 27-minute EuroPython 2018 conference talk. Learn about iterative approaches, avoiding common pitfalls, and maximizing project success using spaCy and Prodigy. Discover techniques for developing effective annotation schemes, model architectures, and pipelines. Gain insights into commercial machine learning project challenges and how to address them through practical examples and workflow recommendations. Understand the importance of evaluation, annotation projects, and iterative processes in building robust NLP solutions.
Overview
Syllabus
Introduction
About spaCy
Why NLP projects fail
How to maximize the risk of NLP projects
Failure sucks
Hierarchy of needs
Circular dependency
iterative process
waterfall approach
Annotation example
Rulebased logic example
General approach
Workflow
Solution
Evaluation
Annotation Projects
The Solution
Taught by
EuroPython Conference
