This readme covers how the necessary dictionaries were generated.

• "Brazil geocodes state, city, municipal n=27, 135, 5572.xlsx" comes from mail thread "Brezilya için geocode, adm2 - 2020".
• "gadm40_BRA_3.json" was generated from "gadm40_BRA_3.dbf" file that comes from "https://www.ibge.gov.br/en/geosciences/full-list-geosciences/18890-municipal-mesh.html?=&t=downloads" 2021.

For district_coords_dict:

• I used "Brazil geocodes state, city, municipal n=27, 135, 5572.xlsx" to generate this dictionary.
• There were some duplicate names in "Municipality name" column, so I added "State name" to these values. For example; "São Domingos" occured 5 times, so for "São Domingos" in "Sergipe" I changed the name to "São Domingos, Sergipe"
• All the Municipality names has locality 6
• Then, I added Intermediarias having locality 3. Most of these names were already in the dictionary as municipality names. I updated these to have locality 3, and updated their coordinates to have intermediarias coordinates.
• "Rio de Janeiro" and "São Paulo" Intermediarias were also state names. I did not add these to state_alternatives

For district_alternatives:

• Don't forget to lower everything
• As usual, I first added the keys in the dictionary as themselves
• Then, I started with "gadm40_BRA_3.json", containing districts under municipalities. I made sure to match the municipalities to our district_coord_dict first.
• I added the alternatives to municipalities from the excel files sheet named "district to add municipalities"
• Dropped the alternatives that were district names themselves but were alternatives to other district names
• Dropped the alternatives that pointed to multiple names (not themselves)
• I added equivelants of some utf-8 characters such as 'é'->'e' as new entries.
• I added unicodedata.normalize() 'd versions of the names.
• For names with state names appended at the end with ', ', I added '()' paranthesis option.
• Finally, went over place_name_frequencies from pipeline output

For state_alternatives:

• Don't forget to lower everything
• As usual, I first added the states as themselves
• I added equivelants of some utf-8 characters such as 'é'->'e' as new entries.
• I added unicodedata.normalize() 'd versions of the names.
• Finally, went over place_name_frequencies from pipeline output