Spaces:
Sleeping
Sleeping
langdonholmes
commited on
Commit
β’
0c29fae
1
Parent(s):
30b20e5
refactor gender country matching
Browse files
app.py
CHANGED
@@ -1,14 +1,17 @@
|
|
1 |
|
2 |
'''Streamlit app for Student Name Detection models.'''
|
3 |
-
from piilo.analyzer import prepare_analyzer
|
4 |
-
from piilo.anonymizer import surrogate_anonymizer
|
5 |
-
import pandas as pd
|
6 |
-
from annotated_text import annotated_text
|
7 |
-
from json import JSONEncoder
|
8 |
import json
|
|
|
9 |
import warnings
|
|
|
|
|
|
|
10 |
import streamlit as st
|
11 |
-
import
|
|
|
|
|
|
|
|
|
12 |
os.environ['TOKENIZERS_PARALLELISM'] = 'false'
|
13 |
warnings.filterwarnings('ignore')
|
14 |
|
@@ -56,7 +59,7 @@ st.set_page_config(page_title='Student Name Detector (English)', layout='wide')
|
|
56 |
|
57 |
# Side bar
|
58 |
st.sidebar.markdown(
|
59 |
-
'''Detect and anonymize PII in text using an [NLP model](https://huggingface.co/langdonholmes/en_student_name_detector) [trained](https://github.com/aialoe/deidentification-pipeline) on student-generated text collected
|
60 |
'''
|
61 |
)
|
62 |
|
@@ -74,7 +77,7 @@ st_return_decision_process = st.sidebar.checkbox(
|
|
74 |
'Add analysis explanations in json')
|
75 |
|
76 |
st.sidebar.info(
|
77 |
-
'This is part of a
|
78 |
)
|
79 |
|
80 |
# Main panel
|
|
|
1 |
|
2 |
'''Streamlit app for Student Name Detection models.'''
|
|
|
|
|
|
|
|
|
|
|
3 |
import json
|
4 |
+
import os
|
5 |
import warnings
|
6 |
+
from json import JSONEncoder
|
7 |
+
|
8 |
+
import pandas as pd
|
9 |
import streamlit as st
|
10 |
+
from annotated_text import annotated_text
|
11 |
+
|
12 |
+
from piilo.engines.analyzer import prepare_analyzer
|
13 |
+
from piilo.engines.anonymizer import surrogate_anonymizer
|
14 |
+
|
15 |
os.environ['TOKENIZERS_PARALLELISM'] = 'false'
|
16 |
warnings.filterwarnings('ignore')
|
17 |
|
|
|
59 |
|
60 |
# Side bar
|
61 |
st.sidebar.markdown(
|
62 |
+
'''Detect and anonymize PII in text using an [NLP model](https://huggingface.co/langdonholmes/en_student_name_detector) [trained](https://github.com/aialoe/deidentification-pipeline) on student-generated text collected from a massive online open-enrollment course.
|
63 |
'''
|
64 |
)
|
65 |
|
|
|
77 |
'Add analysis explanations in json')
|
78 |
|
79 |
st.sidebar.info(
|
80 |
+
'This is part of a project to develop new anonymization systems that are appropriate for student-generated text.'
|
81 |
)
|
82 |
|
83 |
# Main panel
|
piilo/engines/__init__.py
ADDED
File without changes
|
piilo/{analyzer.py β engines/analyzer.py}
RENAMED
File without changes
|
piilo/{anonymizer.py β engines/anonymizer.py}
RENAMED
@@ -10,11 +10,14 @@ from presidio_anonymizer import AnonymizerEngine
|
|
10 |
from presidio_anonymizer.entities import OperatorConfig
|
11 |
from presidio_anonymizer.operators import OperatorType
|
12 |
|
13 |
-
|
|
|
14 |
|
15 |
logger = logging.getLogger('anonymizer')
|
16 |
-
|
17 |
class NameDatabase(NameDataset):
|
|
|
|
|
|
|
18 |
def __init__(self) -> None:
|
19 |
super().__init__()
|
20 |
|
@@ -29,27 +32,31 @@ class NameDatabase(NameDataset):
|
|
29 |
return {'first_name': fn, 'last_name': ln}
|
30 |
|
31 |
def get_gender(self, first_names: str) -> str:
|
32 |
-
'''Return the most frequent gender code for
|
33 |
or None if a match cannot be found.
|
34 |
'''
|
35 |
gender = NameWrapper(self.search(first_names)).gender
|
36 |
return gender if gender else None
|
37 |
|
38 |
def get_country(self, last_names: str) -> str:
|
39 |
-
'''Return the most frequent country code for a
|
40 |
or None if a match cannot be found.
|
41 |
'''
|
42 |
country = NameWrapper(self.search(last_names)).country
|
43 |
return country if country else None
|
44 |
|
45 |
class surrogate_anonymizer(AnonymizerEngine):
|
|
|
|
|
|
|
46 |
def __init__(self):
|
47 |
super().__init__()
|
48 |
self.names_db = NameDatabase()
|
49 |
self.names_df = pd.read_parquet(name_table)
|
50 |
|
51 |
# keep track of names we have seen
|
52 |
-
self.
|
|
|
53 |
|
54 |
def get_random_name(
|
55 |
self,
|
@@ -63,58 +70,74 @@ class surrogate_anonymizer(AnonymizerEngine):
|
|
63 |
returns two rows of the names dataframe
|
64 |
'''
|
65 |
names_view = self.names_df
|
|
|
66 |
if country:
|
67 |
names_view = names_view[names_view['country'] == country]
|
|
|
68 |
if gender:
|
69 |
names_view = names_view[names_view['gender'] == gender]
|
|
|
70 |
if names_view.size < 25:
|
71 |
-
return
|
72 |
-
|
|
|
|
|
73 |
|
74 |
def generate_surrogate(self, original_name: str) -> str:
|
75 |
'''Generate a surrogate name.
|
76 |
'''
|
|
|
77 |
if original_name == 'PII':
|
78 |
# Every time we call this function, Presidio will validate it
|
79 |
# by testing that the function returns a str when the input is
|
80 |
-
# 'PII'.
|
81 |
return 'PII'
|
82 |
|
83 |
-
# If we have seen this name before, return the same surrogate
|
84 |
-
if original_name in self.seen_names:
|
85 |
-
return self.seen_names[original_name]
|
86 |
-
|
87 |
# Use nameparser to split the name
|
88 |
name = HumanName(original_name)
|
|
|
|
|
89 |
|
90 |
-
|
91 |
-
logger.info(f'Gender set to {gender}')
|
92 |
-
country = self.names_db.get_country(name.last) if name.last else None
|
93 |
-
logger.info(f'Country set to {country}')
|
94 |
-
|
95 |
-
surrogate_name = ''
|
96 |
-
|
97 |
-
name_candidates = self.get_random_name(gender=gender, country=country)
|
98 |
-
|
99 |
-
surrogate_name += name_candidates.iloc[0]['first']
|
100 |
-
logger.info(f'First name surrogate is {surrogate_name}')
|
101 |
-
|
102 |
if name.last:
|
103 |
-
|
104 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
105 |
|
106 |
-
logger.info(f'Returning surrogate name {
|
107 |
|
108 |
-
self.
|
|
|
109 |
|
110 |
-
return
|
111 |
|
112 |
def anonymize(
|
113 |
self,
|
114 |
text: str,
|
115 |
analyzer_results: List[RecognizerResult]
|
116 |
):
|
117 |
-
'''Anonymize identified input using Presidio Anonymizer.
|
|
|
118 |
|
119 |
if not text:
|
120 |
return
|
@@ -149,6 +172,8 @@ if __name__ == '__main__':
|
|
149 |
anonymizer = surrogate_anonymizer()
|
150 |
|
151 |
test_names = ['Nora Wang',
|
|
|
|
|
152 |
'MJ',
|
153 |
'',
|
154 |
'(',
|
|
|
10 |
from presidio_anonymizer.entities import OperatorConfig
|
11 |
from presidio_anonymizer.operators import OperatorType
|
12 |
|
13 |
+
data = Path(__file__).parent.parent.parent / 'data'
|
14 |
+
name_table = data / 'ascii_names.parquet'
|
15 |
|
16 |
logger = logging.getLogger('anonymizer')
|
|
|
17 |
class NameDatabase(NameDataset):
|
18 |
+
'''A wrapper around the names_dataset.NameDataset class.
|
19 |
+
'''
|
20 |
+
|
21 |
def __init__(self) -> None:
|
22 |
super().__init__()
|
23 |
|
|
|
32 |
return {'first_name': fn, 'last_name': ln}
|
33 |
|
34 |
def get_gender(self, first_names: str) -> str:
|
35 |
+
'''Return the most frequent gender code for the provided first name,
|
36 |
or None if a match cannot be found.
|
37 |
'''
|
38 |
gender = NameWrapper(self.search(first_names)).gender
|
39 |
return gender if gender else None
|
40 |
|
41 |
def get_country(self, last_names: str) -> str:
|
42 |
+
'''Return the most frequent country code for a the provided last name,
|
43 |
or None if a match cannot be found.
|
44 |
'''
|
45 |
country = NameWrapper(self.search(last_names)).country
|
46 |
return country if country else None
|
47 |
|
48 |
class surrogate_anonymizer(AnonymizerEngine):
|
49 |
+
'''A wrapper around the presidio_anonymizer.AnonymizerEngine class.
|
50 |
+
'''
|
51 |
+
|
52 |
def __init__(self):
|
53 |
super().__init__()
|
54 |
self.names_db = NameDatabase()
|
55 |
self.names_df = pd.read_parquet(name_table)
|
56 |
|
57 |
# keep track of names we have seen
|
58 |
+
self.seen_first_names = dict()
|
59 |
+
self.seen_last_names = dict()
|
60 |
|
61 |
def get_random_name(
|
62 |
self,
|
|
|
70 |
returns two rows of the names dataframe
|
71 |
'''
|
72 |
names_view = self.names_df
|
73 |
+
|
74 |
if country:
|
75 |
names_view = names_view[names_view['country'] == country]
|
76 |
+
|
77 |
if gender:
|
78 |
names_view = names_view[names_view['gender'] == gender]
|
79 |
+
|
80 |
if names_view.size < 25:
|
81 |
+
# If we don't have enough names, just return a random sample
|
82 |
+
return self.names_df.sample(n=1, weights=self.names_df['count'])
|
83 |
+
|
84 |
+
return names_view.sample(n=1, weights=names_view['count'])
|
85 |
|
86 |
def generate_surrogate(self, original_name: str) -> str:
|
87 |
'''Generate a surrogate name.
|
88 |
'''
|
89 |
+
|
90 |
if original_name == 'PII':
|
91 |
# Every time we call this function, Presidio will validate it
|
92 |
# by testing that the function returns a str when the input is
|
93 |
+
# 'PII'. We don't need to run below code in this case.
|
94 |
return 'PII'
|
95 |
|
|
|
|
|
|
|
|
|
96 |
# Use nameparser to split the name
|
97 |
name = HumanName(original_name)
|
98 |
+
new_name = HumanName()
|
99 |
+
gender, country = None, None
|
100 |
|
101 |
+
# First check if we have seen this name before
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
102 |
if name.last:
|
103 |
+
if name.last in self.seen_last_names:
|
104 |
+
new_name.last = self.seen_last_names[name.last]
|
105 |
+
else:
|
106 |
+
# Sample last name, matching country
|
107 |
+
country = self.names_db.get_country(name.last)
|
108 |
+
logger.info(f'Country set to {country}')
|
109 |
+
new_name.last = self.get_random_name(
|
110 |
+
country=country,
|
111 |
+
)['last'].iloc[0]
|
112 |
+
logger.info(f'Last name surrogate is {new_name.last}')
|
113 |
+
|
114 |
+
if name.first:
|
115 |
+
if name.first in self.seen_first_names:
|
116 |
+
new_name.first = self.seen_first_names[name.first]
|
117 |
+
else:
|
118 |
+
# Sample first name matching gender and country, if available.
|
119 |
+
gender = self.names_db.get_gender(name.first)
|
120 |
+
logger.info(f'Gender set to {gender}')
|
121 |
+
new_name.first = self.get_random_name(
|
122 |
+
gender=gender,
|
123 |
+
country=country,
|
124 |
+
)['first'].iloc[0]
|
125 |
+
logger.info(f'First name surrogate is {new_name.first}')
|
126 |
|
127 |
+
logger.info(f'Returning surrogate name {new_name}')
|
128 |
|
129 |
+
self.seen_first_names[name.first] = new_name.first
|
130 |
+
self.seen_last_names[name.last] = new_name.last
|
131 |
|
132 |
+
return str(new_name)
|
133 |
|
134 |
def anonymize(
|
135 |
self,
|
136 |
text: str,
|
137 |
analyzer_results: List[RecognizerResult]
|
138 |
):
|
139 |
+
'''Anonymize identified input using Presidio Anonymizer.
|
140 |
+
'''
|
141 |
|
142 |
if not text:
|
143 |
return
|
|
|
172 |
anonymizer = surrogate_anonymizer()
|
173 |
|
174 |
test_names = ['Nora Wang',
|
175 |
+
'John Williams',
|
176 |
+
'John H. Williams',
|
177 |
'MJ',
|
178 |
'',
|
179 |
'(',
|
main.py β piilo/main.py
RENAMED
@@ -5,31 +5,26 @@ import logging
|
|
5 |
from fastapi import FastAPI
|
6 |
from fastapi.middleware.cors import CORSMiddleware
|
7 |
|
8 |
-
from
|
9 |
-
from
|
10 |
-
from
|
11 |
|
12 |
-
# Define Student Name Detection Model
|
13 |
configuration = {
|
14 |
'nlp_engine_name': 'spacy',
|
15 |
'models': [
|
16 |
{'lang_code': 'en', 'model_name': 'en_student_name_detector'}],
|
17 |
}
|
18 |
|
19 |
-
# set up logger for this module
|
20 |
logger = logging.getLogger('api')
|
21 |
logging.basicConfig(level=logging.INFO)
|
22 |
|
23 |
-
|
24 |
-
logger.info("Loading Presidio Analyzer and Anonymizer")
|
25 |
analyzer = prepare_analyzer(configuration)
|
26 |
anonymizer = surrogate_anonymizer()
|
27 |
-
logger.info("
|
28 |
|
29 |
-
# Initialize FastAPI
|
30 |
app = FastAPI()
|
31 |
|
32 |
-
# Enable CORS
|
33 |
app.add_middleware(
|
34 |
CORSMiddleware,
|
35 |
allow_origins=["*"],
|
@@ -44,8 +39,7 @@ def hello():
|
|
44 |
|
45 |
@app.post("/anonymize")
|
46 |
def anonymize(anon_req: AnonymizeRequest) -> AnonymizeResponse:
|
47 |
-
|
48 |
-
'''
|
49 |
analyzer_result = analyzer.analyze(anon_req.raw_text,
|
50 |
entities=anon_req.entities,
|
51 |
language=anon_req.language,
|
@@ -62,7 +56,6 @@ def anonymize(anon_req: AnonymizeRequest) -> AnonymizeResponse:
|
|
62 |
|
63 |
if __name__ == "__main__":
|
64 |
import os
|
65 |
-
|
66 |
import uvicorn
|
67 |
|
68 |
uvicorn.run(
|
|
|
5 |
from fastapi import FastAPI
|
6 |
from fastapi.middleware.cors import CORSMiddleware
|
7 |
|
8 |
+
from engines.analyzer import prepare_analyzer
|
9 |
+
from engines.anonymizer import surrogate_anonymizer
|
10 |
+
from models.anonymize import AnonymizeRequest, AnonymizeResponse
|
11 |
|
|
|
12 |
configuration = {
|
13 |
'nlp_engine_name': 'spacy',
|
14 |
'models': [
|
15 |
{'lang_code': 'en', 'model_name': 'en_student_name_detector'}],
|
16 |
}
|
17 |
|
|
|
18 |
logger = logging.getLogger('api')
|
19 |
logging.basicConfig(level=logging.INFO)
|
20 |
|
21 |
+
logger.info("Loading Custom Presidio Analyzer and Anonymizer...")
|
|
|
22 |
analyzer = prepare_analyzer(configuration)
|
23 |
anonymizer = surrogate_anonymizer()
|
24 |
+
logger.info("Loading Successful!")
|
25 |
|
|
|
26 |
app = FastAPI()
|
27 |
|
|
|
28 |
app.add_middleware(
|
29 |
CORSMiddleware,
|
30 |
allow_origins=["*"],
|
|
|
39 |
|
40 |
@app.post("/anonymize")
|
41 |
def anonymize(anon_req: AnonymizeRequest) -> AnonymizeResponse:
|
42 |
+
|
|
|
43 |
analyzer_result = analyzer.analyze(anon_req.raw_text,
|
44 |
entities=anon_req.entities,
|
45 |
language=anon_req.language,
|
|
|
56 |
|
57 |
if __name__ == "__main__":
|
58 |
import os
|
|
|
59 |
import uvicorn
|
60 |
|
61 |
uvicorn.run(
|
test_main.py β piilo/test_main.py
RENAMED
File without changes
|