Discover your latent food graph with this 1...
summary
Claims an automated latent food knowledge graph built from behavioral and content data and shows it works. Uses Representation Learning with query2vec, restavek, and fast menu; tooling includes TensorFlow, gensim, fastText, Spark on EMR, Amazon p2 GPUs and an annoy nearest‑neighbor index. Trains on a year of queries and ~60M sessions (140k restaurants, ~10M menu items), one epoch of query data in ~15 minutes, restaurant model in ~1 hour, embedding dim 30. Qualitative results recover cuisine/dish clusters, spelling robustness, and location bias via geo‑hash augmentation.
AI-written summary · about
chapters (27)
- 0:00Introduction
- 1:03GrubHub platform and user flows
- 2:29Unstructured food data chaos
- 3:53Semantic understanding use cases
- 4:56Latent food graph
- 5:18Query Tyvek background
- 6:02Motivation for query expansion
- 7:49Limitations of query expansion
- 9:37Query embeddings from behavior
- 10:18Query Tyvek architecture and training
- 13:09Query Tyvek results
- 17:24Restaurant embeddings with Restavek
- 17:44Restavek model and personalization
- 18:48Word2Vec to Restavek adaptation
- 21:00Distributional hypothesis for restaurants
- 22:06Restavek training and preprocessing
- 23:52Restavek results and visualization
- 27:28Restavek implementation and data flow
- 29:12Menu item embeddings with Fast Menu
- 29:33Content vs behavior trade-offs
- 30:57Fast Menu architecture and benefits
- 32:22Fast Text for semantic matching
- 33:27Location-aware text selection
- 34:49Geohashes for embeddings
- 36:16Fast Menu training and visualization
- 37:42Fast Menu results
- 39:06Conclusion and Q&A
related talks
35mData Science for Dinner: Recommending Recipes
30mGraph Data Science
22mAI-Powered Personalized Discovery with Graph Search and VLM-Enhanced Embeddings
38mKnowledge Graphs : Adding a New Dimension to Vector Search and Generative AI
28mRole of Knowledge Graphs in Mental Health Diagnosis and Cure
25mYak shaving a good place to eat using non negative matrix factorization
35mIs it Food? An Introduction to Machine Learning
06mData Science