Create new boolean fields based on specific terms appearing in a tokenized pandas dataframe-CodePudding

Looping over a list of terms to search for, I need to create a boolean field for each term according to whether or not it is present in a tokenized pandas series.

List of search terms:

terms = ['innovative', 'data', 'rf']

Dataframe:

df = pd.DataFrame(data={'job_description': [['innovative', 'data', 'science'],
                                            ['scientist', 'have', 'a', 'masters'],
                                            ['database', 'rf', 'innovative'],
                                            ['sciencedata', 'data', 'performance']]})

Desired output for df:

                     job_description  innovative   data     rf
0        [innovative, data, science]        True   True  False
1      [scientist, have, a, masters]       False  False  False
2         [database, rf, innovative]        True  False   True
3  [sciencedata, data, performance]        False   True  False

Criteria:

Only exact matches should be replaced (for example, flagging for 'rf' should return True for 'rf' but False for 'performance')
each search term should get it's own field and be concatenated to the original df

What I've tried: Failed, as it creates a boolean for each term in the series:

df['innovative'] = df['job_description'].explode().str.contains(r'innovative').groupby(level=-1).agg(list)

Failed:

df['innovative'] = df['job_description'].str.contains('innovative').astype(int, errors='ignore')

Failed:

df.loc[df['job_description'].str.contains(terms)] = 1

Failed: I tried implementing what was documented here (See if item in each row of pandas series), but could not adapt it to create new fields or flag properly

Thanks for any help you can provide!

CodePudding user response：

Seems like you can use a nested list comprehension to evaluate if each term exists in each row and assign the list to columns in df:

df[terms] = [[any(w==term for w in lst) for term in terms] for lst in df['job_description']]

Output:

                    job_description  innovative   data     rf
0       [innovative, data, science]        True   True  False
1     [scientist, have, a, masters]       False  False  False
2        [database, rf, innovative]        True  False   True
3  [sciencedata, data, performance]       False   True  False

CodePudding user response：

This should be pretty fast:

e = df['job_description'].explode()
new_df[terms] = pd.concat([e.eq(t).rename(t) for t in terms], axis=1).groupby(level=0).any()

Output:

>>> new_df
                    job_description  innovative   data     rf
0       [innovative, data, science]        True   True  False
1     [scientist, have, a, masters]       False  False  False
2        [database, rf, innovative]        True  False   True
3  [sciencedata, data, performance]       False   True  False