0AlphaZero0 commented 4 years ago

Could it be possible to have a score for each location found? Indeed sometimes it could be good to know why some locations are in the results for example : `

import geograpy places = geograpy.get_geoPlace_context(text="This sentence mention UK as country and London as city.") places.countries ['United Kingdom', 'United States', 'Canada'] places.cities ['London'] places = geograpy.get_geoPlace_context(text="Jin Yin-tan Hospital, Wuhan, China.") places.countries ['China', 'Mexico', 'United States'] places.cities ['China'] `

Something like the following score could be interesting : [('United Kingdom',0.99), ('United States',0.56), ('Canada',0.45)]

A score of confidence could help to avoid those results.

WolfgangFahl commented 4 years ago

How would you like to calculate the score? Currently there are a few possible strategies:

use the population as an indicator e.g. Vienna Austria has a higher population as Vienna Illinois
use the combination of fields as an indicator Wuhan China should limit the location as a grid search just in that area

Please note that the disambiguation is currently only possible with the Locator API.

0AlphaZero0 commented 4 years ago

I think a combination of fields should be the best approach.

WolfgangFahl commented 4 years ago

WolfgangFahl commented 3 years ago

52 now addresses this - the default is to order by population. For our own use case we'll use a more sophisticated version and see what the likelyhood of a location is in our context given how often it is already in our corpus.

somnathrakshit / geograpy3

Score for each location? #21

52 now addresses this - the default is to order by population. For our own use case we'll use a more sophisticated version and see what the likelyhood of a location is in our context given how often it is already in our corpus.