alter specific characters in pandas

Question

Background

I have the following df which contains tokenized Text and P_Name columns and is a modification of including word boundary in string modification to be more specific

P_Name = [list(['Steven', 'I','Jobs']),
          list(['A', 'Ma', 'Mary']),
          list(['U','Kar', 'Jacob'])]

Text =  [list(['Steven', 'I','Jobs', 'likes', 'apples','I', 'too']),
        list(['A','i', 'like', 'A', 'lot', 'of','sports','cares', 'A', 'Ma', 'Mary']),
        list(['the','U', 'Kar', 'Jacob', 'what', 'about', 'U', 'huh','?'])]

import pandas as pd
df = pd.DataFrame({'Text' : Text,  
                  'P_ID': [1,2,3], 
                  'P_Name' : P_Name,

                 })
df
   P_ID    P_Name            Text
0   1   [Steven, I, Jobs]   [Steven, I, Jobs, likes, apples, I, too]
1   2   [A, Ma, Mary]       [A, i, like, A, lot, of, sports, cares, A, Ma, Mary]
2   3   [U, Kar, Jacob]     [the, U, Kar, Jacob, what, about, U, huh, ?]

Goal

1) use the name in P_Name to block the corresponding text in the Text column by placing **block**

2) produce a new column New_Text

Tried

From including word boundary in string modification to be more specific

I have modified the code and tried the following

df['New_Text']=[pd.Series(x).replace(dict.fromkeys(y,'**block**') ).str.cat(sep=' ')for x , y in zip(df['Text'],df['P_Name'])]

Which gives close to what I want but not quite since some letters are being inappropriately labeled **block** e.g I in row 0

 P_ID P_Name Text New_Text
    0             [**block**, **block**, **block**, likes, apples, **block**, too]
    1             [**block**, i, like, **block**, lot, of, sports, cares, **block**, **block**, **block**]
    2             [the, **block**, **block**, **block**, what, about, **block**, huh, ?]

Desired Output

P_ID P_Name Text New_Text
    0              [**block**, **block**, **block**, likes, apples, I, too]
    1              [A, i, like, A, lot, of, sports, cares, **block**, **block**, **block**]
    2              [the, **block**, **block**, **block**, what, about, U, huh, ?]

Question

How do I further modify

df['New_Text']=[pd.Series(x).replace(dict.fromkeys(y,'**block**') ).str.cat(sep=' ')for x , y in zip(df['Text'],df['P_Name'])]

or use new code to achieve my desired output?

Answer 1

You want each ordered occurrence of the P_Name sequence in the Text tokens. This can be achieved by iterating over the Text tokens & checking for equality on the entire P_Name tokens:

df["New_Text"] = df["Text"].apply(lambda tokens: tokens.copy())  # copy original tokens
for tokens, name in zip(df["New_Text"], df["P_Name"]):
    for i, token in enumerate(tokens):
        if tokens[i:i + len(name)] == name:
            tokens[i:i + len(name)] = ["**block**"] * len(tokens[i:i + len(name)])

Depending on your use case you might have the untokenised Text (& P_name) available. If so, then substring matching can be done instead, & then perform tokenisation after.