Building a Scalable AI Chatbot #1: From Rules to Deep Learning
Presentation by Max Frenzel, AI researcher, writer, and digital creative. He is the R&D Lead at Bespoke Inc. and author of the bestselling book Time Off A Practical Guide to Building Your Rest Ethic and Finding Success Without the Stress. Max has also been interested in the applications of AI and deep learning to creativity, design, and music, and he is a regular public speaker on topics such as AI and creativity. In his time off, Max enjoys good coffee, tries to perfect his bread baking skills, and produces electronic music and performs around Tokyo. You can find him online at www.maxfrenzel.com.
Find code and presentation slides here: https://github.com/bespoke-inc/bespok...
---
In this workshop, we will learn how to make a retrieval-based chatbot in a modular fashion using an array of techniques from using rules with regular expressions to more modern deep learning methods such as using large pre-trained language models. We will look at practical challenges, maintaining conversation context, correcting user typos, and extracting parameters using named entity recognition.
We will use sci-kit-learn, PyTorch, fastai, transformers, and spaCy libraries. Participants are expected to have some experience with these tools.
This tutorial will follow a top-down approach. We will start by looking at using higher-level APIs to quickly prototype a solution and slowly dig deeper into the inner workings.
=========================
MLT (Machine Learning Tokyo)
site:
github: https://github.com/Machine-Learning-T...
slack: https://machinelearningtokyo.slack.co...
discuss: https://discuss.mltokyo.ai/
twitter: / __mlt__
meetup: https://www.meetup.com/Machine-Learni...
facebook: / machinelearningtokyo