Recruiter emailed me a detailed assignment: Problem I want to target ads at Twitter users. I’d like to start by identifying tweets which reflect Purchase Intent. I have access to the Twitter firehose. For each tweet with Purchase Intent, I would like to infer the specific item to which such intent applies. Examples Luke Sargent @lsargent02 "Still waiting for the A7s I want a camera that is small enough for me to carry around all the time.” Purchase Intent: High; Item: camera Dani @DaniCarbert17 "Well, lost my phone so if anyone needs to get a hold of me inbox me or in an emergency text Joce LOL” Purchase Intent: Medium; Item: phone Your solution Describe the algorithms, process, etc. for inspecting tweets and classifying those with Purchase Intent. What is the pipeline for ingesting, filtering, classifying, analyzing (and any other steps that come to mind) incoming tweets? Suggest tools that could be used for each of these steps. Make suggestions to maximize scalability & throughput How would you design visualizations so that management can make an informed decision on whether it is worth investing in an advertising tool which targets Twitter users who express Purchase Intent? This could include things like percentage of tweets with PI, specific products which are more likely to be purchased, etc.
