Current status and additional requirements for places inspection & labeling tool

[wagnerfe]

Current status:

• General: the current version accepts a .parquet file, loads previous labeling progress (if existent), and determines next pairs for inspection. Based on current id, it creates an html displaying a map with the places pairs and a table with the respective places attributes
• Display: for each place (base-id, purple) one or several places (can_id, blue) can be displayed. If several places (can_id) are shown, one can select a place in the table using the mouse and or the arrow keys and a yellow circle is displayed on the map highlighting the location of the selected place. Under the hood, we also load the map for the next candidate as well to speed up performance of displaying the map.
• Labeling: via the buttons or with the arrow keys (right arrow:= match, left:arrow = no match) one can label if a place (base_id) is a match with the currently selected place (can_id). Here, one can also jump to the next places pair without labeling using "unsure" or by pressing enter. The labelling progress is stored under data/labeling-progress.pickle and eventually output as results.csv file.

Further Requirements:

Minimal requirements:

1. Display of user-defined columns instead of fixed columns. Currently it requires input data to have the columns name, address, country, house_number, normalized_phone, id with prefix "can_" and "base".
2. Allow moving to next places-pairs w.o. labeling via pressing "enter" in the case of having several places (can_id) for one place (base_id)
3. Allow adding a (or selecting a predefined) category when labeling matches.
4. Display clusters which only contain can_ids (TBD: discuss how labeling could work here or if adding a category label to a cluster would also be sufficient)

Additional requirements:
TBD