Food, Wine and Machine Learning: Teaching a Bot to Taste
summary
Builds a chatbot called Happy Wine Bot that recommends wines for recipes by scraping online recipes, hand-labeling 50+ food features, and classifying text per flavor; wine knowledge uses a codified set of 100+ wine types with 40+ taste attributes. Uses Ruby, integrates Wit.ai for intent parsing and Facebook/Twitter for delivery. Pairing uses a weighted scoring engine (rules rather than ML) to rank matches; demo shows workable recommendations but manual labeling, limited training data and misspellings limit accuracy.
AI-written summary · about
chapters (29)
- 0:00Intro
- 0:54About Mai
- 3:00What is Machine Learning
- 4:03How Machine Learning Works
- 4:55What Can Machine Learning Do
- 7:00Why arent you using Machine Learning
- 8:50Cloud Services
- 10:47Data
- 12:17What makes good data
- 13:37Feature engineering
- 15:47Data management
- 17:31Predicting
- 17:50When not to use machine learning
- 19:09Challenges with machine learning
- 22:01Practical example
- 22:59Teaching the bot to converse
- 24:25What is taste
- 25:36Features of food that affect taste
- 26:18Training data
- 27:18Models
- 28:13Classification Example
- 28:58Features of Wine
- 29:59Teaching the Bot
- 31:51Sommelier Strategies
- 34:02How to Teach the Bot
- 34:42Demo
- 36:18Wine Recommendations
- 37:25Lessons Learned
- 38:58Takeaway
related talks
35mData Science for Dinner: Recommending Recipes
54mMachine Learning explained with Examples
33mA Sommelier's Guide to Recommendation Algorithms: Classical and Graph-Based Recommender Systems
43mDeveloper Accessible Machine Learning
44mArtificial Intelligence and Machine Learning in the browser
48mAn App That Sees, Hears and Responds
35mHow we trained a bot to understand fashion requests
55mDemystifying Natural Language Processing across several industry verticals