Ideation

Steve Klabnik: WASM + Rust + Serverless

Rust

WASM

Serverless

Rasa NLU Lessons

  1. Train INTENT and MESSAGE in single NN to be language agnostic (any lang tokenizable)
  2. Out-of-data and Fallback policies to improve service
  3. Form collection required from user before Decision
  4. Feedback loop of system itself logging to improve the system (i.e. logging user-bot conversation to extract more labels)

spaCy Conference Talks

Mark Neumann: SciSpacy and Non-ML Extension

Build Domain Model

  1. Entities
  2. POS + Dependencies
  3. Vector
  4. Mixing in generic text to have both worlds

spaCy IRL slide

Quartz Language Model

link to slide


code · notebook · prose · gallery · qui et quoi? · main