A simple HTTP endpoint in Python using Flask that predicts sentiment for a given sentence
(sentiment analysis). The service authenticates the user (found in users_DB.txt), give the opportunity to select a model and
return one of the 3 sentiments: “happy”, “neutral”, or “unhappy”.
GET request handler for the route /predict.
-
Input: /predict?username=dwight&model=happy&sentence=i-am-very-happy
-
Output: minimal json response: { “predicted_sentiment”: "happy" }
-
Input: /predict?username=john&model=length-based&sentence=i-am-very-happy-but-my-sentence-is-long
-
Output: minimal json response: { “predicted_sentiment”: "unhappy" }
In programmatic advertising, ads (commonly referred to as impressions) are shown on websites with the help of DSPs (Demand Side Platforms). These platforms want to make sure that they are showing the right impressions to the right users in order to generate as many conversions as possible. Conversion is an event that occurs when a user lands on the advertiser's website. Put simply - when a user opens a website which contains some space for an ad, a request is sent to our servers where choose the right impression, which we serve to the user. We also store the request data in a table called impressions. If a user clicks on the ad and lands on the advertiser's website we store this event in a table called conversions. For each conversion there exists at least one impression, but not all impressions have a conversion.