通过比较替换 python 中列表/数据框列中的字符串

Question

I have the following dataframe column

Hotels
Hotel Tulsi Viha
Hotel Tulsi Vih
Hotel Tulsi Vihar
SWIGGYBang
Swiggy
Borivali Biryani center
Borivali Biryani centr

I want to check for string similarity and replace it:- For eg Hotel Tulsi Vih, Hotel Tulsi Viha, Hotel Tulsi Vihar are same, so I want string "Hotel Tulsi Vihar" that replaces the other 2 strings. Similarly it needs to be done on entire column

Answer 1

You can apply fuzzywuzzy library to check the similar string.

For example, you have a main list

from fuzzywuzzy import process, fuzz

main_list = ['Hotel Tulsi Vihar','Swiggy','Borivali Biryani center']

## compare the similar string to a list, as below
process.extract("Hotel Tulsi Viha", main_list , limit=2, scorer=fuzz.token_sort_ratio)

## Result -> [('Hotel Tulsi Vihar', 97), ('Swiggy', 18), ('Borivali Biryani center', 15)]

So, you can apply this function to a dataframe

## Create/Import dataframe
df = pd.DataFrame(['Hotels','Hotel Tulsi Viha','Hotel Tulsi Vih','Hotel Tulsi Vihar'\
,'SWIGGYBang','Swiggy','Borivali Biryani center','Borivali Biryani centr'],columns=['Brand'])

## This is example function to compare word with a list and threshold. 
def check_fuzzywuzzy(word, list_unique,threshold):
    highest_pos = process.extract(word, list_unique, limit=1, scorer=fuzz.token_sort_ratio)
    if highest_pos[0][1] > threshold:
        return highest_pos[0][0]
    else:
        return word

## Apply function
df['new_column'] = df.apply(lambda x: check_fuzzywuzzy(x['Brand'], main_list,60), axis=1)

# Result
#                    Brand          new_column_name
#0                   Hotels                   Hotels
#1         Hotel Tulsi Viha        Hotel Tulsi Vihar
#2          Hotel Tulsi Vih        Hotel Tulsi Vihar
#3        Hotel Tulsi Vihar        Hotel Tulsi Vihar
#4               SWIGGYBang                   Swiggy
#5                   Swiggy                   Swiggy
#6  Borivali Biryani center  Borivali Biryani center
#7   Borivali Biryani centr  Borivali Biryani center

Then, you can use the new column, or you can apply to the same column, 'Brand'.

For the fuzzywuzzy library, it has several score for checking the similar string, eg, token_set_ratio, partial_ratio. More information is in this link .