Spaces:
Running
Running
langdonholmes
commited on
Commit
·
3ad7899
1
Parent(s):
e37bcc3
refactor anonymizer with inheritance
Browse files- anonymizer.py +115 -56
- app.py +4 -5
- names_database.py +21 -28
anonymizer.py
CHANGED
@@ -1,67 +1,111 @@
|
|
1 |
-
|
|
|
|
|
2 |
|
|
|
3 |
from presidio_analyzer import RecognizerResult
|
4 |
from presidio_anonymizer import AnonymizerEngine
|
5 |
from presidio_anonymizer.entities import OperatorConfig
|
|
|
6 |
|
7 |
from names_database import NameDatabase
|
8 |
|
9 |
-
|
10 |
|
11 |
-
|
12 |
-
'''Splits name into parts.
|
13 |
-
If one token, assume it is a first name.
|
14 |
-
If two tokens, first and last name.
|
15 |
-
If three tokens, one first name and two last names.
|
16 |
-
If four tokens, two first names and two last names.'''
|
17 |
-
names = original_name.split()
|
18 |
-
if len(names) == 1:
|
19 |
-
return names[0], None
|
20 |
-
elif len(names) == 2:
|
21 |
-
return names[0], names[1]
|
22 |
-
elif len(names) == 3:
|
23 |
-
return names[0], ' '.join(names[1:])
|
24 |
-
elif len(names) == 4:
|
25 |
-
return ' '.join(names[:2]), ' '.join(names[2:])
|
26 |
-
else:
|
27 |
-
return None, None
|
28 |
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
46 |
|
47 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
48 |
|
49 |
-
def anonymize(
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
|
59 |
-
|
60 |
-
|
61 |
-
|
62 |
-
|
|
|
|
|
63 |
'STUDENT': OperatorConfig('custom',
|
64 |
-
{'lambda': generate_surrogate}),
|
65 |
'EMAIL_ADDRESS': OperatorConfig('replace',
|
66 |
{'new_value': '[email protected]'}),
|
67 |
'PHONE_NUMBER': OperatorConfig('replace',
|
@@ -69,9 +113,24 @@ def anonymize(
|
|
69 |
'URL': OperatorConfig('replace',
|
70 |
{'new_value': 'aol.com'}),
|
71 |
}
|
72 |
-
|
73 |
-
|
74 |
-
|
|
|
|
|
|
|
|
|
|
|
75 |
|
76 |
if __name__ == '__main__':
|
77 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import logging
|
2 |
+
from pathlib import Path
|
3 |
+
from typing import List, Optional
|
4 |
|
5 |
+
import pandas as pd
|
6 |
from presidio_analyzer import RecognizerResult
|
7 |
from presidio_anonymizer import AnonymizerEngine
|
8 |
from presidio_anonymizer.entities import OperatorConfig
|
9 |
+
from presidio_anonymizer.operators import OperatorType
|
10 |
|
11 |
from names_database import NameDatabase
|
12 |
|
13 |
+
name_table = Path('data', 'ascii_names.parquet')
|
14 |
|
15 |
+
logger = logging.getLogger('anonymizer')
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
16 |
|
17 |
+
|
18 |
+
class surrogate_anonymizer(AnonymizerEngine):
|
19 |
+
def __init__(self):
|
20 |
+
super().__init__()
|
21 |
+
self.names_db = NameDatabase()
|
22 |
+
self.names_df = pd.read_parquet(name_table)
|
23 |
+
|
24 |
+
|
25 |
+
def get_random_name(
|
26 |
+
self,
|
27 |
+
country: Optional[str] = None,
|
28 |
+
gender: Optional[str] = None
|
29 |
+
) -> pd.DataFrame:
|
30 |
+
'''Returns two random names from the database as a DataFrame.
|
31 |
+
Both rows match gender and country, if provided.
|
32 |
+
:country: ISO country code e.g. "CO" for Columbia
|
33 |
+
:gender: 'M' or 'F'
|
34 |
+
returns two rows of the names dataframe
|
35 |
+
'''
|
36 |
+
names_view = self.names_df
|
37 |
+
if country:
|
38 |
+
names_view = names_view[names_view['country'] == country]
|
39 |
+
if gender:
|
40 |
+
names_view = names_view[names_view['gender'] == gender]
|
41 |
+
if names_view.size < 25:
|
42 |
+
return self.names_df.sample(n=2, weights=self.names_df['count'])
|
43 |
+
return names_view.sample(n=2, weights=names_view['count'])
|
44 |
+
|
45 |
+
def split_name(self, original_name: str):
|
46 |
+
'''Splits name into parts.
|
47 |
+
If one token, assume it is a first name.
|
48 |
+
If two tokens, first and last name.
|
49 |
+
If three tokens, one first name and two last names.
|
50 |
+
If four tokens, two first names and two last names.'''
|
51 |
+
names = original_name.split()
|
52 |
+
if len(names) == 1:
|
53 |
+
logger.info(f'Splitting to 1 first name: {names}')
|
54 |
+
return names[0], None
|
55 |
+
elif len(names) == 2:
|
56 |
+
logger.info(f'Splitting to 1 first name, 1 last name: {names}')
|
57 |
+
return names[0], names[1]
|
58 |
+
elif len(names) == 3:
|
59 |
+
logger.info(f'Splitting to 1 first name, 2 last names: {names}')
|
60 |
+
return names[0], ' '.join(names[1:])
|
61 |
+
elif len(names) == 4:
|
62 |
+
logger.info(f'Splitting to 2 first names and 2 last names: {names}')
|
63 |
+
return ' '.join(names[:2]), ' '.join(names[2:])
|
64 |
+
else:
|
65 |
+
logger.info(f'Splitting failed, do not match gender/country: {names}')
|
66 |
+
return None, None
|
67 |
+
|
68 |
+
def generate_surrogate(self, original_name: str):
|
69 |
+
'''Generate a surrogate name.
|
70 |
+
'''
|
71 |
+
first_names, last_names = self.split_name(original_name)
|
72 |
+
gender = self.names_db.get_gender(first_names) if first_names else None
|
73 |
+
logger.debug(f'Gender set to {gender}')
|
74 |
+
country = self.names_db.get_country(last_names) if last_names else None
|
75 |
+
logger.debug(f'Country set to {country}')
|
76 |
+
|
77 |
+
surrogate_name = ''
|
78 |
|
79 |
+
name_candidates = self.get_random_name(gender=gender, country=country)
|
80 |
+
|
81 |
+
surrogate_name += name_candidates.iloc[0]['first']
|
82 |
+
logger.info(f'First name surrogate is {surrogate_name}')
|
83 |
+
|
84 |
+
if last_names:
|
85 |
+
logger.info(f'Combining with {name_candidates.iloc[1]["last"]}')
|
86 |
+
surrogate_name += ' ' + name_candidates.iloc[1]['last']
|
87 |
+
|
88 |
+
logger.info(f'Returning surrogate name {surrogate_name}')
|
89 |
+
return surrogate_name
|
90 |
|
91 |
+
def anonymize(
|
92 |
+
self,
|
93 |
+
text: str,
|
94 |
+
analyzer_results: List[RecognizerResult]
|
95 |
+
):
|
96 |
+
'''Anonymize identified input using Presidio Anonymizer.'''
|
97 |
+
|
98 |
+
if not text:
|
99 |
+
return
|
100 |
+
|
101 |
+
analyzer_results = self._remove_conflicts_and_get_text_manipulation_data(
|
102 |
+
analyzer_results
|
103 |
+
)
|
104 |
+
|
105 |
+
operators = self._AnonymizerEngine__check_or_add_default_operator(
|
106 |
+
{
|
107 |
'STUDENT': OperatorConfig('custom',
|
108 |
+
{'lambda': self.generate_surrogate}),
|
109 |
'EMAIL_ADDRESS': OperatorConfig('replace',
|
110 |
{'new_value': '[email protected]'}),
|
111 |
'PHONE_NUMBER': OperatorConfig('replace',
|
|
|
113 |
'URL': OperatorConfig('replace',
|
114 |
{'new_value': 'aol.com'}),
|
115 |
}
|
116 |
+
)
|
117 |
+
|
118 |
+
res = self._operate(text,
|
119 |
+
analyzer_results,
|
120 |
+
operators,
|
121 |
+
OperatorType.Anonymize)
|
122 |
+
|
123 |
+
return res.text
|
124 |
|
125 |
if __name__ == '__main__':
|
126 |
+
logging.basicConfig(level=logging.DEBUG)
|
127 |
+
anonymizer = surrogate_anonymizer()
|
128 |
+
test_names = ['Nora Wang',
|
129 |
+
'MJ',
|
130 |
+
'',
|
131 |
+
'(',
|
132 |
+
'Mario Escobar Sanchez',
|
133 |
+
'Jane Fonda Michelle Rousseau',
|
134 |
+
'Sir Phillipe Ricardo de la Sota Mayor']
|
135 |
+
for name in test_names:
|
136 |
+
anonymizer.generate_surrogate(name)
|
app.py
CHANGED
@@ -2,7 +2,7 @@
|
|
2 |
'''Streamlit app for Student Name Detection models.'''
|
3 |
|
4 |
from analyzer import prepare_analyzer
|
5 |
-
from anonymizer import
|
6 |
from presidio_anonymizer import AnonymizerEngine
|
7 |
import pandas as pd
|
8 |
from annotated_text import annotated_text
|
@@ -31,8 +31,8 @@ def analyzer_engine():
|
|
31 |
|
32 |
@st.cache(allow_output_mutation=True)
|
33 |
def anonymizer_engine():
|
34 |
-
'''Return
|
35 |
-
return
|
36 |
|
37 |
def annotate(text, st_analyze_results, st_entities):
|
38 |
tokens = []
|
@@ -116,10 +116,9 @@ with st.spinner('Analyzing...'):
|
|
116 |
st.text('')
|
117 |
|
118 |
st.subheader('Anonymized')
|
119 |
-
|
120 |
with st.spinner('Anonymizing...'):
|
121 |
if button or st.session_state.first_load:
|
122 |
-
st_anonymize_results =
|
123 |
st_text,
|
124 |
st_analyze_results)
|
125 |
st_anonymize_results
|
|
|
2 |
'''Streamlit app for Student Name Detection models.'''
|
3 |
|
4 |
from analyzer import prepare_analyzer
|
5 |
+
from anonymizer import surrogate_anonymizer
|
6 |
from presidio_anonymizer import AnonymizerEngine
|
7 |
import pandas as pd
|
8 |
from annotated_text import annotated_text
|
|
|
31 |
|
32 |
@st.cache(allow_output_mutation=True)
|
33 |
def anonymizer_engine():
|
34 |
+
'''Return generate surrogate anonymizer.'''
|
35 |
+
return surrogate_anonymizer()
|
36 |
|
37 |
def annotate(text, st_analyze_results, st_entities):
|
38 |
tokens = []
|
|
|
116 |
st.text('')
|
117 |
|
118 |
st.subheader('Anonymized')
|
|
|
119 |
with st.spinner('Anonymizing...'):
|
120 |
if button or st.session_state.first_load:
|
121 |
+
st_anonymize_results = anonymizer_engine().anonymize(
|
122 |
st_text,
|
123 |
st_analyze_results)
|
124 |
st_anonymize_results
|
names_database.py
CHANGED
@@ -1,42 +1,35 @@
|
|
1 |
-
|
2 |
-
|
3 |
|
4 |
-
import pandas as pd
|
5 |
from names_dataset import NameDataset, NameWrapper
|
6 |
|
7 |
-
name_table = Path('data', 'ascii_names.parquet')
|
8 |
|
9 |
class NameDatabase(NameDataset):
|
10 |
def __init__(self) -> None:
|
11 |
super().__init__()
|
12 |
-
|
13 |
-
|
14 |
-
def get_random_name(
|
15 |
-
self,
|
16 |
-
country: Optional[str] = None,
|
17 |
-
gender: Optional[str] = None
|
18 |
-
):
|
19 |
-
'''country: ISO country code in 'alpha 2' format
|
20 |
-
gender: 'M' or 'F'
|
21 |
-
returns two rows of the names dataframe
|
22 |
-
'''
|
23 |
-
names_view = self.names
|
24 |
-
if country:
|
25 |
-
names_view = names_view[names_view['country'] == country]
|
26 |
-
if gender:
|
27 |
-
names_view = names_view[names_view['gender'] == gender]
|
28 |
-
if names_view.size < 25:
|
29 |
-
return self.names.sample(n=2, weights=self.names['count'])
|
30 |
-
return names_view.sample(n=2, weights=names_view['count'])
|
31 |
|
32 |
-
def search(self, name: str):
|
|
|
|
|
|
|
|
|
33 |
key = name.strip().title()
|
34 |
fn = self.first_names.get(key) if self.first_names is not None else None
|
35 |
ln = self.last_names.get(key) if self.last_names is not None else None
|
36 |
return {'first_name': fn, 'last_name': ln}
|
37 |
|
38 |
-
def get_gender(self, first_names: str):
|
39 |
-
|
|
|
|
|
|
|
|
|
40 |
|
41 |
-
def get_country(self, last_names: str):
|
42 |
-
|
|
|
|
|
|
|
|
|
|
1 |
+
import logging
|
2 |
+
|
3 |
|
|
|
4 |
from names_dataset import NameDataset, NameWrapper
|
5 |
|
|
|
6 |
|
7 |
class NameDatabase(NameDataset):
|
8 |
def __init__(self) -> None:
|
9 |
super().__init__()
|
10 |
+
|
11 |
+
self.logger = logging.getLogger('anonymizer')
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
|
13 |
+
def search(self, name: str) -> dict:
|
14 |
+
'''Returns all entries associated with a name string.
|
15 |
+
The name string can be multiple tokens.
|
16 |
+
Both first and last names will be matched.
|
17 |
+
'''
|
18 |
key = name.strip().title()
|
19 |
fn = self.first_names.get(key) if self.first_names is not None else None
|
20 |
ln = self.last_names.get(key) if self.last_names is not None else None
|
21 |
return {'first_name': fn, 'last_name': ln}
|
22 |
|
23 |
+
def get_gender(self, first_names: str) -> str:
|
24 |
+
'''Return the most frequent gender code for a specific last name,
|
25 |
+
or None if a match cannot be found.
|
26 |
+
'''
|
27 |
+
gender = NameWrapper(self.search(first_names)).gender
|
28 |
+
return gender if gender else None
|
29 |
|
30 |
+
def get_country(self, last_names: str) -> str:
|
31 |
+
'''Return the most frequent country code for a specific last name,
|
32 |
+
or None if a match cannot be found.
|
33 |
+
'''
|
34 |
+
country = NameWrapper(self.search(last_names)).country
|
35 |
+
return country if country else None
|