Building State-of-the-art Natural Language Processing Projects with Free Software - Scale by the Bay 2021
- 1. Building Complete State-of-the-art Natural Language Processing Projects with Free Software David Talby CTO, John Snow Labs
- 2. 2 Agenda Introduction Python’s NLU Library The Annotation Lab The NLP Server 1. 2. 3. 4.
- 3. 60% of Technical Leaders stated that their NLP budget was at least 10% higher vs. 2020 33% of Technical Leaders stated that their NLP budget was at least 30% higher vs. 2020 NLP is Growing Into a Foundational Enterprise Technology NLP Industry Survey 2021, Gradient Flow, July-August 2021 “Half of Technical Leaders are using technologies for data integration, NLP, business intelligence, and data warehousing.” Healthcare AI Survey 2021, Gradient Flow, January-March 2021
- 4. SO, HOW DO WE BEST HELP THE COMMUNITY & INDUSTRY SUCCEED?
- 5. BY GIVING YOU MORE FREE SOFTWARE (THAT IS PROVEN, ROBUST, SCALABLE, AND SECURE)
- 6. 6 BUILDING ON SPARK NLP >10 million O’Reilly Media 31% share of NLP Practitioners use Spark NLP Gradient Flow 24x growth In downloads of the library during 2020 PyPI Download Stats Downloads. Most popular NLP tool in the enterprise
- 7. 7 Agenda Introduction Python’s NLU Library The Annotation Lab The NLP Server 1. 2. 3. 4.
- 8. INTRODUCING THE NLU LIBRARY The Simplicity of Python, the Power of Spark NLP
- 9. EXAMPLE: TOKENIZATION & SPELL CHECKING
- 15. MULTI-CLASS TOXIC CONTENT CLASSIFIER
- 16. MULTI-CLASS TOXIC CONTENT CLASSIFIER
- 17. SUMMARIZATION
- 19. NLU ALSO HAS ONE-LINER VISUALIZATIONS
- 21. NLU EVEN HAS ONE-LINER STREAMLIT APPS
- 22. NLU STREAMLIT APPS CHEAT SHEET
- 23. Getting Started with the NLU Library ONE LINE INSTALLATION: DOCUMENTATION: https://nlu.johnsnowlabs.com/docs COMMUNITY SUPPORT: https://spark-nlp.slack.com @ #nlu
- 24. 24 Agenda Introduction Python’s NLU Library The Annotation Lab The NLP Server 1. 2. 3. 4.
- 25. JOHN SNOW LABS’ ANNOTATION LAB IS NOW FREE Free Unlimited users, documents, projects, tasks, models, … Share nothing: Self hosted on your infrastructure Robust 2 years in production Now in its 23rd version Actively developed & supported Capable Active learning, Transfer learning Model training & deployment Projects & Teams Roles & Workflows Versioning & Audit Analytics: Productivity, Agreement, Bias, Teamwork
- 26. High Productivity Annotation Label entities, relations, assertions, classes, values, sequences, … Keyboard Shortcuts Versioned & Auditable Completions Start from model predictions Shared Guidelines One-click edits
- 27. Teamwork
- 28. Projects & Roles
- 29. Pre-Annotate: Don’t Start from Scratch Filter Selected Tasks Choose Model & Embedding Filter Selected Model Labels Asynchronous Model Server (Runs on Kubernetes)
- 30. Automated Model Training & Active Learning Async Training Container Realtime Training Log Filter Completions by State or Tag Edit Hyper-Parameters Choose Embeddings Train & Deploy Or Fully Automate
- 31. Download, Upload, and Publish Models
- 32. Analytics Project Tracking Productivity Bias in Dataset or Guidelines
- 33. Inter-Annotator Agreement Agreement between annotators, or between human & model Drill-down to specific cases
- 34. Getting Started with the Annotation Lab ONE LINE INSTALLATION: DOCUMENTATION: https://nlp.johnsnowlabs.com/docs/en/alab/quickstart COMMUNITY SUPPORT: https://spark-nlp.slack.com @ #annotation-lab
- 35. 35 Agenda Introduction Python’s NLU Library The Annotation Lab The NLP Server 1. 2. 3. 4.
- 36. JOHN SNOW LABS’ NLP SERVER IS NOW FREE Free Unlimited models, calls, servers Share nothing: Self hosted on your infrastructure Early Stage Started in 2021 Now in its 3rd version Actively developed & supported Capable No coding required 3,000+ models & growing UI + REST API
- 37. The User Interface 1. Search spells by name or description 2. Type text to analyze or upload a CSV 3. Hit the ‘Preview’ button
- 38. Viewing Results 1. Models are downloaded to local storage 2. Models are loaded & cached in memory 3. Model inference is executed
- 40. The REST API Same functionality as the user interface, delivered as a single Docker container
- 41. NLP Server Roadmap Model training & auto-tuning Serve your custom trained NLP models Run on air-gap enterprise environments Model publishing from the Annotation Lab Run Healthcare NLP & Spark OCR models
- 42. Getting Started with the NLP Server ONE LINE INSTALLATION: DOCUMENTATION: https://nlp.johnsnowlabs.com/docs/en/nlp_server/nlp_server COMMUNITY SUPPORT: https://spark-nlp.slack.com @ #nlp-server
- 43. Please put it all to good use! https://nlp.johnsnowlabs.com/docs
- 44. 44 Thank you! © 2015-2021 John Snow Labs Inc. All rights reserved. The John Snow Labs logo is a trademarks of John Snow Labs Inc. The included information is for informational purposes only and represents the current view of John Snow Labs as of the date of this presentation. Since John Snow Labs must respond to changing market conditions, it should not be interpreted to be a commitment on its part, and John Snow Labs cannot guarantee the accuracy of any information provided after the date of this presentation. John Snow Labs makes no warranties, express or statutory, as to the information in this presentation. david@JohnSnowLabs.com @davidtalby in/davidtalby