From 0 to a crowd sourced neural network in a jiffy
summary
A working production-grade NLP pipeline is built on stage: Doccano annotation feeds an RDS-backed dataset that is packaged to S3, sent to an AWS GPU training API, fine-tuned via Hugging Face on a multilingual BERT model, and deployed to a prediction API that serves live NER and an anonymize API. Models trained on as few as ~100 labeled reviews show measurable F1 gains within minutes; the live demo intermittently fails on venue Wi‑Fi but the system architecture, tools, and end-to-end workflow are fully implemented.
AI-written summary · about
chapters (9)
related talks
43mDetecting fake reviews at scale using Apache Spark and John Snow Labs
49mMachine Learning 2.0 With Hugging Face
17mLaying the foundation for deep learning
01h 25mBuild and Deploy an End-to-end Streaming NLP Insight System
45mArchitecture of an NLP Deployment
30mProductionizing an unsupervised machine learning model to...
01h 30mFrom Synthetic Examples to Production Signals
30mBuilding Complete State of the art Natural Language Processing