Classify a stream of tokens with embeddings

Given data that arrives as a stream, process the tokens online, turn them into embeddings, and build a model that classifies them. Cover how you process data as it arrives, which embeddings you use, the classifier, latency, and how you keep the model current.