ETNA: Harvesting Action Graphs from Websites
summary
Etna extracts action graphs from websites by turning interaction traces into timelines, classifying fine-grained states (empty vs populated cart), and producing generalized action functions with input parameters. It mines collections of UI snapshots to create parameterizable deeplinks and UI scripts, and when deeplinks fail it uses semantic ranking to robustly locate elements for replay. Evaluations show automated replay and state classification work across many commercial sites, and Etna powers generated APIs, zero-code chatbots (reservations, weather, movies), RPA improvements, and NLP datasets.
AI-written summary · about
related talks
38mData and Business
16mEvolution of Extract-Transform-Load (ETL) processes towards data product pipelines
18m1,200 analytics requests a day: Clay and Vercel's Bet on AI-Native Analytics
26mHigh Performance Data Mart Using Event Plots
03mInterview
26mGrafana at Continental Scale: Monitoring the National Ecological Observatory Network (NEON)
45mManaging Data & Analytics Projects Effectively
26mEthereum Blockchain Analytics