Htykuut

I'm trying to replace some adresses in French in a dataframe. I'm using a list and regular expressions and one list.

def adresses(df):  



    liste_adresses = ['allée', 'Allée', 'rue', 'Rue', 'avenue', 'Avenue', 'av', 'AV', 'boulevard', 'Boulevard', 'bd', 'Bd', 'carreau', 'Carreau', 'carrefour', 'Carrefour', 'place', 'Place', 'voie', 'Voie', 'villa', 'Villa', 'route', 'Route', 'quai', 'Quai']



    for i in liste_adresses:



        df['C'] = df['C'].str.replace(r'[0-9]+(,|s+)is+w+s+(w+)?(s+)?(w+)?(s+)?([0-9]{5})?(s+)?w+?([0-9]{5})?','<address>')



return df

My dataframe:

       A          B                                                                C

  French      house                      I live in 15 rue Louis Philippe 75001 Neuilly

 English      house               my address: 101-102 bd Charles de Gaulle 75001 Paris

  French  apartment                                                    my name is Liam

  French      house                                                       Hello George!

 English  apartment  This is wrong: 4, rue Ledion Paris 75014 and I'm not happy with it

On my output, nothing happens.

Good output:

       A          B                         C

  French      house                                I live in <address>

 English      house                              my address: <address>

  French  apartment                                    my name is Liam

  French      house                                       Hello George!

 English  apartment  This is wrong: <address> and I'm not happy with it

The following solution may not works for specific cases. Because the end of the address is either the postal code or the city that you don't know, I think one way could be to look for:

1. a string with numbers at first '[0-9]+': all addresses start with a number


2. some characters (.*): for example to catch -102
   


3. any word from liste_adresses using '|'.join(liste_adresses)
   


4. the postal code of 5 digits [0-9]{5}
   


5. look for the city name if existing with ([^.|n]{0,2}[A-Z][a-z]*)*: here I assume that if there is a dot or a new line after the postal code, then the address is over, so match between 0 and 2 characters but not a dot or new line [^.|n]{0,2}, then one upper case letter [A-Z] then any lower case [a-z]* until the end of the word, the extra at the end * would catch cities composed of two words like Saint-Denis.

So globally, doing:

liste_adresses = ['allée', 'Allée', 'rue', 'Rue', 'avenue', 'Avenue', 'av', 'AV',

                  'boulevard', 'Boulevard', 'bd', 'Bd', 'carreau', 'Carreau',

                  'carrefour', 'Carrefour', 'place', 'Place', 'voie', 'Voie',

                  'villa', 'Villa', 'route', 'Route', 'quai', 'Quai']



reg = r'[0-9]+(.*)('+'|'.join(liste_adresses) + ')(.*)[0-9]{5}([^.|n]{0,2}[A-Z][a-z]*)*'



print (df['C'].str.replace(reg,'<address>'))

0                                  I live in <address>

1                                my address: <address>

2                                      my name is Liam

3                                        Hello George!

4    This is wrong: <address> and I'm not happy wit...