You can use a Python library like fuzzywuzzy
here, which has support for this type of task:
from fuzzywuzzy import process
df.assign(Output=[process.extract(i, df['Col-1'], limit=3) for i in df['Col-2']])
Using the process
method, we can get string similary scores, and then pick the top 3, if 3 exist:
The output of the above code:
Col-1 Col-2 Output
0 Update have a account [(Account, 90, 1), (AccountDTH, 64, 2), (Update, 40, 0)]
1 Account account summary [(Account, 90, 1), (AccountDTH, 63, 2), (Credit Card, 38, 4)]
2 AccountDTH Cancel [(Balance, 62, 3), (Credit Card, 43, 4), (Update, 33, 0)]
3 Balance Balance Summary [(Balance, 90, 3), (Credit Card, 38, 4), (Update, 30, 0)]
4 Credit Card Update credit card [(Update, 90, 0), (Credit Card, 90, 4), (AccountDTH, 27, 2)]
To speed this comparison up (natively it uses Python's sequence matcher), I would recommend installing python-Levenshtein