The book Human-in-the Loop Machine Learning by Randall Munro has a code example of finding hazards in food safety reports. Here's the description of the problem: Food Safety professionals want to collect data from incident reports about where pathogens or foreign objects have been detected in food. “I want to maintain a complete record of all recorded food safety incidents in the EU” “I want to track when different food safety incidents might have come from the same source” “I want to send warnings to specific countries when there are likely to be food safety incidents that have not yet been detected or reported” The interface has fields for "Hazard", "Food", "Origin", and "Destination" along with a short extract of text from a food report.
The book Human-in-the Loop Machine Learning by Randall Munro has a code example of annotating bicycles. Here's the description of the problem: Transportation researchers want to estimate the number of people who use bicycles on certain streets. “I want to collect information about how often people are cycling down a street” “I want to capture this information from thousands of cameras and I don’t have the budget to do this manually” “I want my model to be as accurate as possible” Based on this he has designed an interface for rapid annotation.
The book Human-in-the Loop Machine Learning by Randall Munro has a code example of annotating headlines for a data analyst. First you choose a topic name and then can annotate examples. For example I chose "sports results" intending to label headlines containing the result of a sport contest (and not other kinds, e.g. political contests). It wasn't totally obvious how to annotate examples at first; I had to click in the box with the example headline.
I've been reading Robert Monarch's Human-in-the-Loop Machine Learning and the second chapter has a great practical example of human-in-the-loop machine learning; identifying whether a news headline is about a disaster. A lot of the Data Science books, courses, and competitions take for granted that you've got a well defined machine learning metric to optimise, and labelled dataset. In practice you often have a practical objective, and have to define the metric and collect the input features and labels.