Next Gen Restaurant Recommendation
- Source
- uber.com
- Author
- Uber Engineering 2
- Date
Context
Uber Eats can now rank restaurants using what a customer did earlier in the same session. Uber Engineering says in its account of the redesign that its models used to depend on batch features computed by offline jobs, with a lag of 24 hours or more, and that the lag is now a few seconds.
The system keeps a near-real-time, event-based history of a user's actions and computes features from it on the fly. A -based sequence model reads the chronological log of clicks and orders and, in what Uber calls target-aware modeling, appends the restaurant being scored to the sequence, so the model can relate past behavior to that particular candidate. The resulting feeds the main ranking model. Uber describes the change as particularly helpful for users with little or no history, but the article gives no figures for that effect.
- transformer — The neural network architecture behind modern AI models, built on attention — letting every word directly consider every other word in parallel.
- embedding — A list of numbers representing a piece of text's meaning, so that similar meanings end up numerically close and can be searched.
- attention — The mechanism that lets a model weigh which earlier words matter for the word it's currently processing — the core operation of a transformer.
Checking sign-in…
Loading comments…