PyCon Colombia Speaker Interview

2020-02-08 · Source: Explosion · Developer tools and consulting for AI, Machine Learning and NLP - Explosion.ai · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Data Science & Analytics · Depth: Intermediate, long

Summary

Ines Montani, co-founder of Explosion and core developer of spaCy and Prodigy, shared her journey into NLP and the motivations behind her company's flagship products. Starting programming at age 11, Montani later merged her linguistics and media science background with coding, leading her to NLP. spaCy originated from co-founder Matthew Honnibal's observation of a market gap for a production-ready NLP library, distinct from research-oriented tools. Explosion then developed Prodigy, an annotation tool, after identifying critical data quality and labeling challenges during consulting projects. Prodigy's success has funded Explosion, allowing it to remain independent and support open-source development. Looking ahead, Explosion aims to empower users to build highly customized, lower-level models, integrating diverse ML frameworks like TensorFlow and PyTorch into spaCy for advanced transfer learning and data labeling. Montani also emphasized the importance of deep domain expertise, suggesting machine learning should serve as an additive skill for solving real-world problems.

Key takeaway

For NLP Engineers and Data Scientists building production-ready systems, prioritize robust, efficient tooling like spaCy and invest in high-quality, iterative data annotation with tools such as Prodigy. Your efforts should focus on understanding the problem domain deeply, as machine learning serves best as an additive skill to specialized knowledge. Consider leveraging Explosion's future capabilities to integrate custom models and diverse ML frameworks for tailored solutions.

Key insights

Production-ready NLP tools and quality data annotation are crucial, with domain expertise being more vital than ML for its own sake.

Principles

• Production-focused libraries fill a critical ecosystem gap.
• High-quality data annotation improves model performance.
• Domain expertise is foundational for effective ML application.

Method

Prodigy enables iterative workflows: try an idea, label data, understand it, and refine the model for better results.

In practice

• Use spaCy for fast, efficient NLP processing.
• Employ Prodigy for iterative, high-quality data labeling.
• Integrate custom models with spaCy using Thinc's functional API.

Topics

• Natural Language Processing
• spaCy
• Prodigy
• Data Annotation
• Machine Learning Engineering
• Domain Expertise

Best for: NLP Engineer, Machine Learning Engineer, Director of AI/ML

Related on AIssential

• A Bit of AI Episode 7 — AI developer tools like spaCy and Prodigy are crucial for democratizing and streamlining NLP solution development.
• How to build resilient NLP applications — Production-grade NLP requires embracing inherent complexity, providing granular control, and integrating domain expertise for real-world value.

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.