Al 14c89e6895 [addresses] utilities for sampling from an arbitrary discrete distribution, building cumulative distributions, and sampling from a Zipfian distribution which seems to be a reasonable way of generating plausible apartment/floor numbers when the height/number of units is unknown. Picking a letter uniformly at random means P('Unit A') == P('Unit Z') when 'A' should be much more likely. Sampling from a Zipfian gets the desired effect in situations where address components are numbered by "counting from 0/1/A" while still allowing for a long tail

2016-07-21 17:04:57 -04:00

..

__init__.py

[addresses] address config class for general sampling of forms specified in the address configs (default/alternatives to choose a phrase, canonical/abbreviated/sample to choose an abbreviation or surface form for that phrase)

2016-07-21 17:04:57 -04:00

config.py

[addresses] address config class for general sampling of forms specified in the address configs (default/alternatives to choose a phrase, canonical/abbreviated/sample to choose an abbreviation or surface form for that phrase)

2016-07-21 17:04:57 -04:00

sampling.py

[addresses] utilities for sampling from an arbitrary discrete distribution, building cumulative distributions, and sampling from a Zipfian distribution which seems to be a reasonable way of generating plausible apartment/floor numbers when the height/number of units is unknown. Picking a letter uniformly at random means P('Unit A') == P('Unit Z') when 'A' should be much more likely. Sampling from a Zipfian gets the desired effect in situations where address components are numbered by "counting from 0/1/A" while still allowing for a long tail

2016-07-21 17:04:57 -04:00