spaCy v3: State-of-the-art NLP from Prototype to Production

2021-02-01 · Source: Explosion · Developer tools and consulting for AI, Machine Learning and NLP - Explosion.ai · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, extended

Summary

spaCy v3, a widely adopted open-source Python NLP library, introduces significant enhancements for production-grade natural language processing. The release features new transformer-based pipelines, leveraging Huggingface Transformers and PyTorch, which achieve state-of-the-art accuracy, notably reducing NER errors by 30%. A novel configuration system, built on Python's configparser with JSON values and registered functions, streamlines complex ML setups, offering Pydantic-based validation and CLI integration for training and hyperparameter tuning. Furthermore, spaCy v3 includes a Projects system for managing multi-stage ML workflows via YAML templates, supporting remote caching and integrations with tools like DVC and Weights & Biases. The extensible pipeline system, powered by the Thinc library, allows seamless integration and customization of models from various frameworks, while improved developer experience is delivered through Python type hints and Thinc's model type annotations.

Key takeaway

For NLP Engineers building production systems, spaCy v3 significantly simplifies complex workflows and boosts model performance. You should explore its transformer-based pipelines for state-of-the-art accuracy and leverage the new configuration system for robust, validated model setups. Adopt spaCy Projects to standardize and share multi-stage ML workflows, ensuring easier deployment and collaboration. This release empowers you to move from prototype to production with greater efficiency and reliability.

Key insights

spaCy v3 streamlines production NLP with transformer integration, robust configuration, and structured project management.

Principles

• Production NLP demands robust, deployable software.
• Transformers offer superior scaling and accuracy.
• Bottom-up configuration enhances system flexibility.

Method

spaCy Projects use YAML templates to define assets, commands, and workflows, enabling dependency tracking, remote caching via spacy project push/pull, and easy replication of multi-stage ML tasks.

In practice

• Utilize transformer pipelines for improved accuracy.
• Define custom components using Language.factory.
• Manage multi-stage ML projects with spaCy Projects.

Topics

• spaCy
• NLP Pipelines
• Transformers
• ML Workflow Management
• Configuration Systems
• Python NLP

Best for: AI Architect, MLOps Engineer, NLP Engineer, Machine Learning Engineer, AI Engineer

Related on AIssential

• Introducing spaCy v3.0 — spaCy v3.0 achieves state-of-the-art NLP accuracy via transformer pipelines and streamlines project workflows from prototype to production.
• spaCy v3: State-of-the-art NLP from Prototype to Production — spaCy v3 delivers production-ready NLP by integrating Transformers, a robust config system, and project management for scalable, highly accurate applications.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Explosion · Developer tools and consulting for AI, Machine Learning and NLP - Explosion.ai.